Organizing Safe Spaces: #MeToo Activism in Sweden

Networked online environments can effectively support political activism. In Sweden, the #metoo movement resulted in over 100,000 people participating in activities challenging sexual harassment and abuse, including collecting testimonies via social media and drafting and discussing petitions published in print news media. Participation involved many risks, such as social stigma, losing one’s job, or misogynist terrorism, which meant that participation required a high level of trust among peers. Human-computer interaction (HCI) research on trust generally focuses on technical systems or user-generated data, less focus has been given to trust among peers in vulnerable communities. This study, based on semi-structured interviews and surveys of participants and organizers of 47 petitions representing different sectors in society, found that trust was aggregated over networks of people, practices, institutions, shared values, and technical systems. Although a supportive culture based on a feeling of solidarity and shared feminist values was central for safe spaces for participation, when activism was scaled up, social interaction had to be limited due to increased risk. HCI research views trust as a process of crossing distances, increasing over time; however, our results reveal that trust decreased over time as the movement grew and public exposure increased, a trend most evident when the participants actually came from a tightly knit community. Therefore, this study points out the significance to balance the need for transparency and community with the need for anonymity and distance in the development of tools to support large-scale deliberative processes that involve conflicts and risks.


Introduction
Networked online environments can effectively serve as settings for the organization and mobilization of social movements. Previous research shows how social media for example has been used to organize and engage the public in the environmental movement (DeLuca et al. 2016;Goodwin and Jasper 2014;Pang and Law 2017), the Arab Spring (AlSayyad and Guvenc 2015; Smidi and Shahin 2017), the Occupy Movement (Kavada 2015), the Gezi protests in Turkey (Haciyakupoglu and Zhang of being a target of sexual abuse. In addition, participants feared becoming a target of threats or continued harassment and hate crimes. Because of these risks the trust in the movement was crucial, to make participants willing to publicly share their experiences of traumas. In this paper, we therefore seek to understand the organization of the #metoo activism with a focus on trust. A shorter first version of this article was published as an exploratory paper in (Hansson et al. 2019). In comparison, this article is substantially developed and incorporates an analysis of a larger dataset involving not only organizers of the movement but also participants.

Trust
Trust is a central concept in human-computer interaction (HCI) research. When navigating the topic of trust online, research mainly investigates e-commerce solutions (Bauman and Bachmann 2017;Cheng et al. 2017;Corritore et al. 2003;Kracher et al. 2005), e-government systems (Bannister and Connolly 2011;Le Dantec 2018a, 2018b), and e-health systems (Beldad et al. 2010). The focus in these areas is mainly on how consumers and citizens can feel confident in systems that handle sensitive data such as money or medical records (Beldad et al. 2010;Lampinen and Cheshire 2016;Wang and Emurian 2005). When it comes to trust in people, the focus has often been on the relationship between the citizen/ consumer and the authority/service and therefore not directly about the trust between peers (Corbett and Le Dantec 2018a). For example, based on a student survey, Lankton and Harrison McKnight (2011) came to the conclusion that trust in Facebook is about trust in the technology as well as in the social network it represents. Similarly, in their ethnography of the uses of Facebook and the mass media in Tunisia, Aal et al. (2018) show the importance of the discursive process in social media for enabling trust. When it comes to situations with higher risks, such as when sharing resources in a neighborhood (Light and Miskelly 2019), or sharing information in a crisis situation (Tapia and Moore 2014), trust is foremost in people and not the technology. However, as research on communications in the Syrian civil war points out, while one's social network is central for trust in information, the technology enables a relatively intact infrastructure in the crisis. In addition, Moser et al. (2017) explain the trust mechanisms that enable transactions between strangers in some Facebook groups, where trust is fostered through exclusive membership of a closed group, moderation by the administrator, and a shared group identity based on perceived similar values (rather than on social bonding).
Another relevant aspect of trust in online settings is personal safety. The relation between the desire for self-exposure and the possibility of being anonymous has been demonstrated when sensitive subjects are addressed (Birnholtz et al. 2015), vulnerable groups such as targets of sexual abuse are exposed (Andalibi et al. 2016), and women who have had a miscarriage are identified (Andalibi and Forte 2018). At the same time, research on people's safety awareness on social media shows that even though there are concerns that sensitive information is being revealed, the benefits of sharing experiences and gaining support are perceived as so valuable that it outweighs the risks. This also applies to vulnerable groups such as illegal immigrants in the United States (Guberek et al. 2018). An important strategy to build a trusted network is separatism, often used in feminist and queer activism for creating safe spaces (Clark-Parsons 2018;Scheuerman et al. 2018;Sills et al. 2016). Separatism provides environments where one's experiences and identity can be recognized by likeminded but also for exchanging and developing ideas which in the long run can influence the general publicwhat Fraser (1990) calls subaltern counter-publics.
Undoubtedly, the extensive HCI research on trust shows the centrality and complexity of trust within the field. It is also a multifaceted concept that means different things in different contexts, and there is no consensus on how to approach the concept, or how to measure it (Grabner-Kräuter et al. 2006;Söllner et al. 2016;Söllner 2020). For the sake of clarity, in the following we therefore describe how we define and use the concept of trust in this paper. This definition is also informed by our informants' ideas about trust, risk and safety. Haraway (1991) defines technology as a kind of prosthesis that allows us to stretch our "arms" beyond our bodies and reach what we previously could not reach. In this view, trust is about trusting that our arms can reach what we are targeting and carry what we expect them to do. There is always a risk that the prosthesis will break, but most of the time it goes well. Risk implies that trust is required, so risk and trust are closely associated: the more risk, the more trust required.
Trust is also linked to distancethe greater the distance, the greater the trust required. The concept of distance includes physical distance, temporal distance, emotional distance, and social distance (Corbett and Le Dantec 2018a). Here, trust can be seen as a process of bridging distances, a process that can be described in various phases such as developing, building, and maintaining trust (Ibid). In the development phase, trust is about a calculated and weak confidence. Trust in this phase is mainly cognitive and is about relying on clear evidence and strong external structures such as laws and systems. In the construction phase, trust is more about experience built through interactions over time where previously trusted people and situations are trusted again. The third phase, maintaining trust, is less about calculations and more about belonging and takes its point of departure in shared values and benevolence. That is, during the maintaining trust phase, people trust not only that the system will work and people are predictable, but also that people are motivated by shared values.
Typically, social and cultural distances decrease as people get to know each other, by e.g. meeting face to face (Zheng et al. 2001), or solving problems together in the neighborhood (Light and Miskelly 2019). However, this may be an oversimplified picture, as getting close to each other may also give rise to, or make visible, conflicts between various interests. The participants in the Swedish #metoo movement did not always share views on the reasons and solutions to the problem of sexual harassment, and they also disagreed on other questions. Still, the Swedish #metoo movement succeeded in uniting large groups of people around the issue. A useful term for describing this is Spivak's (1988)strategic essentialism, denoting a political subject whose identity position is temporary, contingent and conditioned by an emancipatory political goal. The #metoo activism was thus not only risky for external reasons, but internal inequalities also entailed potential risks.

Data and methods
To understand how the organizers of the #metoo petitions went about to create safe spaces, this study uses a mixed-methods approach, consisting of eight semi-structured interviews with organizers of petition groups and two surveys with organizers and participants, where the results of the interviews guided the design of the surveys. The preliminary results have been shared with the informants, both as a means to clarify any misunderstandings and to spur further discussion. By examining the factors that influence participation, we not only investigate the conditions for activism, but also support the participants' self-reflection and thus provide tools for continued activism. This article thus contributes to the area of feminist HCI, as it is about sharing feminist activism but also adheres to a methodology where the research is firmly grounded in informants' perspective, focusing on supporting participatory emancipatory processes (Bardzell and Bardzell 2011).

Semi-structured interviews
The eight informants were selected to represent a diverse group of professional and interest groups from a number of contexts: the IT and construction industries, the green industry, the arts, and sports. As most of the organizers used e-mail pseudonyms or were not publicly known, we did not know their age or background.
Each interview lasted between 50 and 70 min and began with a brief overview of the purpose of the research, followed by a series of questions asking the informants to describe their background and role in the organization of the petition, what ideas and values influenced the organization, how the petition was organized, how it was distributed, what role safety and trust played, and what they had learned from the experience.

Surveys
The first survey was distributed to the organizers of all 77 written petitions initiated between November 2017 and June 2018 (Hansson 2020). The number of contact persons and how they were contacted varied. Some groups provided group aliases that transferred e-mail to all the organizers of the petition, and others provided individual addresses of one or a few of the organizers. Some petitions were organized by groups of people, and others by just one person. The petitions differed in reach as well: some collected over 10,000 signatures and some collected far fewer signatures, mainly from a closely-knit group. Of the 105 e-mail and Facebook Messenger addresses that were sent a survey, 62 organizers of 47 petitions responded.
The second survey was addressed to participants in two of the petitions, #skiljaagnarnafrånvetet (#separatethewheatfromthechaff) representing the green industry and #sistaspikenikistan (#lastnailinthecoffin) representing the construction industry. The choice of these specific groups was made to get a sample of two professional areas that differed in many aspects. Their original Facebook groups were still intact, so we could distribute the survey to the same groups of people that had participated in the petition. An open call in the two Facebook groups resulted in the response from 56 participants from all over Sweden, who completed the survey. For safety reasons, we do not distinguish different informants based on which petition they organized or participated in. Instead we aggregate them in role and type of study, and differentiate them based on letters and numbers; interview with organizers (A-E), survey with organizers (1-62), and survey with participants (1-56).
The surveys asked similar questions as those asked in the interviews; about how the petitions were organized, what tools and methods were used, motivation, safety, and lessons learned. The interviews were recorded and transcribed. As all data were in Swedish, the quotations have been translated into English.
The interviews, as well as the open-ended questions in the survey, were analyzed thematically, where an open coding of the data was followed by a more focused coding to identify salient categories and organize the material in different themes. The initial coding was made by the main author and developed together with the coauthors.
The material used in this article is on the one hand the respondents' thoughts on communication tools and processes, as expressed in the surveys, and on the other hand on the dominating themes from the analysis of the open survey answers and the interviews that we have named strategies for creating safe spaces.

Results
The petitions were initiated by individuals or smaller groups who formulated the petition texts, and then engaged a larger collective in collecting and discussing testimonies and collecting signatures. When the petition was formulated, it was distributed to via these collectives, as well as to newspapers, radio, and television.
The results from the interviews showed that the organizers were between 25 and 54 years old. Their previous experiences of organizing activism were mixed, from no experience at all to long experience in political activism, for example through union work. Before becoming an organizer of a petition group, many had access to some sort of professional network online. For example, some already served as moderators for social media groups gathering people from their industry or were responsible for e-mail lists that connected former classmates, thus there was already a technical affordance in place, with access to communication tools and social networks.
The 62 respondents of the survey addressed to the organizers represented 47 petitions: from smaller areas such as comedians (signed by 80 persons) to larger ones such as physicians (signed by 10,400 persons) (see full list in Table 1 in Appendix).
The organizers came from all over Sweden, from Malmö in the south to Kiruna in the north. One lived in Finland, 31 in Stockholm, and four in Gothenburg, the second largest city in Sweden. The remaining 27 respondents came from various small towns or rural areas. The organizers were between 20 and 69 years old with the majority (44 of 62) being between 30 and 49 years old. 54 had a college education, which is twice the average of the general Swedish population (SCB 2018).
The 56 participants in the survey addressed to participants in the petition-groups from the green industry and the construction industry came from all over Sweden. Their ages were quite evenly distributed, between 25 and 74 years old. Educational levels were average for Sweden (28 had a college education).

Communication tools and processes
According to the survey with organizers, the tools for initiating and developing the petitions varied from meetings to phone calls, e-mails, and collaborative writing to social media and survey tools. In most cases, social media meant Facebook, which was used by all petition groups, in combination with other tools. E-mail or text messaging platforms were used in half of the cases, and Google docs was used in one-third of the petition groups. Twitter and Instagram were sparsely used and mainly to complement Facebook as a way to distribute the petitions. This shows how Facebook dominated in all phases of communication, but also how various tools were used in combination.
The smaller group of organizers typically used a group on Facebook, a text messaging platform, or text chat as an exclusive channel to communicate among themselves. Most organizing groups maintained a close and continual contact through different tools. To communicate with the participants they used a combination of methods and tools, and the communication processes varied. In the initiating phase, social media were the most important tools, and according to the participants in our second survey, most had first learned about the petition through social media.
#metoo created strong affect among many people in society. Suddenly, people talked about sexual harassment among friends, around the coffee tables, and in countless newspaper articles. Organizing and participating in a petition channeled these feelings. Main reasons given for organizing as well as participating in petitions was showing solidarity with victims, as well as having a strong feminist conviction. For example, one petition organizer said that she took the initiative to the petition as an angry reaction to a male colleague who claimed that their industry had no problems with sexual harassment.
The first published petitions inspired other professional or interest groups to initiate activity. For example, #teknisktfel (technical error), which was the petition of the tech industry, started as a discussion thread in an already existing Facebook group that gathered thousands of women from the industry. When news media extended the issue beyond famous actors by publishing #metoo petitions of other professions and interest groups, the issue was discussed in a discussion thread in this group and several participants asked for a petition in their own industry. In response, one of the participants in short time wrote a manifesto and set up a survey tool to collect signatures online, and she also created an e-mail address that others could send their testimonies to.
Interviewer: How did the petition start? Did you know each other before starting the petition?
Respondent: No, we did not. There was a then newly formed group on Facebook. It grew quite explosively in the fall and there were a few different threads, and when these petitions came, it was #tystnadtagning (#silencetaking) first, and then the singers' petition came and then the third which was the lawyers. And that was, to me, such a punch in the stomach, because then what made that petition differ from the others was that now it was no longer famous people who took advantage of their media space and voice, but it was ordinary people who came together and gathered. And, of course, it became a huge discussion as well, and several people said that this is something that is a problem in our industry too. And it was late, it was 8-9 in the evening when it started to be discussed, and everyone said, 'someone should do it (a petition)', 'Can you? And then they pointed to someone else. And then it became 11 o'clock and no one did anything, so I pressed the button, thinking I might just as well do the collection of signatures then. I googled "collection of signatures" on the internet, and so I found this namninsamling.se (name collection) where you could gather names. And then I wrote a text, only spontaneously, and then I spread it. (Interview with organizer A) The informant quoted above posted information about the petition in an open Facebook group, which meant that the petition quickly received a wide distribution. After a few hours, she had enough material to draft an article proposal aimed at a leading newspaper. The whole process from idea to draft went very fast, taking no more than 4-5 h ( Figure 1). To get help developing and completing the article, she asked for help from the others in the discussion thread, forming a group around the continued work. For this group, it was thus a shared strong affect, accessible infrastructure, and access to a network of over 15,000 women in the same industry that altogether made the activism successful.
Other petitions developed much slower, taking their energy from life-long frustration. Often, these testimonies were published in semi-public or closed Facebook groups, which generated long discussions on each case. In some petitions, the text was developed collectively, not only by the organizers but all members of the group were invited to come with suggestions, and the text was examined in detail and discussed intensively before publishing. Several of these cases took place in Facebook groups with thousands of participants. Other petitions avoided social media for internal communication and used it only as a means to distribute the petition. #slutavverkat (#stopthefelling), representing the forest industry, was foremost published via an Instagram account that published 162 testimonies one-by-one between December 2017 and April 2018 ( Figure 2).
According to the interviews with organizers, the later petition groups learned from the experiences of previous ones, for example by making organizers more cautious about how they used social media or whether they published their private e-mail addresses. Prior to this experience, most participants had not experienced what it meant to be in the public eye. The social media groups that had seemed to be safe semi-private rooms began to leak information and expand to become more insecure public spheres.
After the initiating phase, testimonies were collected through e-mail, social media, and survey tools. One-fourth of the petitions used some sort of survey tool to collect  (2) was used to collect signatures and a manifesto was drafted on google docs (3) and discussed on Facebook. After a few hours, an article proposal was submitted to the leading daily newspaper (4) and published a few days later. The article was widely distributed on social media (1) testimonies and signatures; the others used either a Facebook group or a dedicated email list.
In the final distribution phase, the petition was published in leading news media, industry publications, and websites ( Figure 3). Especially the first month, the two largest daily newspapers (Dagens Nyheter and Svenska Dagbladet), and the major tabloid (Aftonbladet) were important for the wide distribution of the petitions. These were also the newspapers with the highest perceived legitimacy. After November, industry publications and television became more important for the distribution of petitions. In further communication with stakeholders, social media were central, but seminars and meetings with journalists and decision makers were also important for promoting and exchanging ideas.

Strategies for creating a safe space
Although the whole idea of #metoo was to make sexual harassment visible and decrease the shame of having been exposed, the public exposure put the participants at great risk for retaliation. For example, disclosures could lead to unpleasant consequences both socially and professionally in the form of threats, harassment, and social exclusion. Legally, accusations that cannot be substantiated in concrete evidence or other witnesses could lead to the person reporting the case being prosecuted for defamation (on the other hand, perpetrators who were publicly identified could suffer from extreme consequences not proportional to the alleged crime). Therefore, it is not surprising that one central theme in the open questions in the surveys and in the interviews dealt with strategies for creating safe spaces.
The material suggests five overarching strategies. The first and most dominant strategy emphasizes a supportive community and openness, and the other four strategies emphasize regulation and control.

Supportive communities
Our results confirm previous research showing that even in risky situations, sharing experiences and support in a community is often perceived as so valuable that it outweighs the risks (Andalibi et al. 2016;Andalibi and Forte 2018;Birnholtz et al. 2015;Guberek et al. 2018). Most petition organizers emphasized the importance of a trustful environment that encouraged participants to talk about their experiences and to receive support from women with similar experiences. To achieve this, it was central to the organization of the petitions to establish trustful forms of dialogue. Through active and collective moderating, a supportive and generous culture was The situation was based on the shared feeling of affect, a generalized trust in shared feminist values, and confidence in the organizers and their ability to harbor trust.
Most stories were submitted to me and the other organizer. Some published their stories in the Facebook group, which created trust so that other people also dared to share. This in turn created more trust. But sending by e-mail felt safer and we were careful to ask before posting the stories that this really was ok. I think that we showed great respect from the beginning and that this was a good start. We as organizers set rules for what we could talk about and not in the group.
I believe the safety of these groups is largely based on the evidence of how widespread the problem is. If, for the first time, you feel that you are listened to and taken seriously and if you sympathize with others in the group, then the interests to break the social rules is not so great.
(Survey answer from organizer 39) Most participants in our survey (51 of 56) did not share testimonies, but mostly participated by sharing information, signing the petition and distributing the petition in their networks. As the informant above explains, in some of the Facebook groups, participants shared their testimonies directly, and their identity was known by everyone. The other participants supported the testifiers with "likes" and supporting comments, which encouraged others to contribute their experiences. Thus, being recognized and seeing others contribute, along with the transparency of the groups, developed another kind of trust, a trust in the community of collective experience. When testimonies were met by support and feedback from a large group, and led others into sharing their experiences, the contribution of stories of traumatic experiences became meaningful.
Sometimes, reading about what others had experienced and discussing the prevalence and nature of sexual harassment could make participants remember repressed experiences and reevaluate normalized sexist behavior. The discussion groups thus functioned as deliberative spaces, where insights were shared and developed. As mentioned above, organizers emphasized their explicit ambition to create a supporting and generous environment for discussion, without judging or blaming. This supportive culture was further reproduced by the participants, for example by giving many likes and supportive comments to participants who shared experiences of being exposed to sexual harassment or violence. Solidarity with the movement per se was also the motivation most often expressed by the participants. The groups were formed out of a shared strong affect, and through the supportive culture, anger and sadness was channeled into active support. This was further supported by the Facebook interface, which, through its design, encouraged supportive feedback. However, considering the fast pace of the process, discussions could develop for a couple of days or weeks but seldom longer, which in most cases did not leave room for developing any lasting relationships. The trust was thus more a generalized trust in shared values and practices than trust in specific persons or communities.
Several of the organizers had a readiness to handle people who needed more support, for example by providing information about people or organizations that provide legal or psychological support.
We as administrators and coordinators took an active role and set the tone in the comment fields. There was never hatred or bullshit; instead, many pointed out how good the ambiance was. The focus was on 'Thank you for telling us' and always reminding you that there was the opportunity to get more support. We worked a lot with responsiveness and, for example, using a language that did not exclude. From the very beginning, we created an opportunity for anyone who wanted to talk to a person in charge at our federal office if they needed more support and/or wanted to report a perpetrator to possibly move on in some way. (Survey answer from organizer 22) The organizers were subjected to significant pressure. Often with no experience of organizing, they suddenly found themselves in charge of moderating intense discussions on social media groups with over 1000 participants, while also communicating with journalists and key actors in their field, organizing seminars and participating in public discussions. Thus, there was a need for a supportive environment also among organizers. In the surveys, we received many similar answers to how a trusting environment was created, emphasizing a communicative strategy where all means were used to maintain close contact with the group, to encourage each other and to share experiences: -Continuous communication between us, we met a lot and talked a lot about what was sent to us. (Organizer 34) -Mainly through active contact and support between us. We replaced each other when there were tougher discussions in our Facebook group (Organizer 42).
-We who organized were in different places so could not meet physically but had close contact over Messenger so we would always be on the same wavelength. (Organizer 44) -We had our own WhatsApp group where we supported and pepped each other all the time. (Organizer 55) -A lot of conversations, support and a constant checking with each other.
(Organizer 56) Some organizers knew each other before the organization of the petition and these relationships facilitated communication. However, most did not know one another before the petition. One important factor that influenced the degree of trust, among participants and organizers alike, was therefore the identity of organizers. Several petitions were led by women who were well known in the industry and thereby c r e d i bl e s p o k e s p e r s on s . F o r e x a m p l e , on e o f t h e or g a n i z e r s o f #skiljaagnarnafrånvetet (#separatethewheatfromthechaff) was a well-known person in the green industry, both as a lecturer and as a writer of a chronicle in the industry magazine. She had previously initiated a women's network on Facebook to support her in the development of a book about the conditions for women in the field, and this network became instrumental when the petition was initiated. Another example is #sistaspikenikistan (#lastnailinthecoffin), where one of the organizers had previously worked with questions of gender equality at the construction industry union Byggnads(Building) and could therefore easily reach out to union activists and create a wider engagement for the question.

Separatism
Central to the organization of the Swedish #metoo petition groups was creating spaces to which potential perpetrators, or people who opposed the movement, had no access. Such subaltern counter-publics have a long feminist tradition, as separatist spaces enabled the formulation of feminist politics (Fraser 1990). For the #metoo petitions, social media provided an effective platform, as the organizers invited appropriate participants from their networks, or posted open invitations in established feminist fora. However, even if participants were recruited through social media or e-mail, most often by someone they knew, this did not mean that they knew all others in the group, as these could consist of several thousand people.
One way to ensure that information did not seep out of the group was to carefully check and limit new members, so that no unauthorized persons gained access to the group. This could sometimes stir up conflict: Here is the crux [. . .] it was decided, for example, that no journalists would be allowed to join the group. Then part of the admin group went in with the argument 'but that's my friend' and added these people again. Also industry professionals/ service persons at [industry name] were added with the same argument against the group's will, even though the group assumed to be a group for [professional identity]. Thus, they were expected to tell sensitive stories to their superiors, who in some cases leaked information to [the industry company]. (Survey answer from organizer 16) Many times, questions about power imbalance and dependency conditions could be a dilemma, which sometimes made the organizers take other paths, and use tools other than Facebook: We didn't want to bring in some of the people we knew about [who had high positions or were responsible for staff]. So, we never arranged such a [Facebook group], but instead we spread the Google form via Messenger and yes, we sent it to our nearest network, and so it spread. So then it became like one could pass it around and say that it comes from a safe source. It was as if we passed the trust on.

(Interview with organizer E)
For most petitions, the question of who could participate was simple: Women in the industry in question. For many petitions "women" meant "women or non-binary people". In several cases, however, discussions arose about the question of who would be allowed to participate. The least controversial question was the separatism, the exclusion of men, as including men could lead to the presence of potential perpetrators, which would reduce the trust within the group. In industries where career paths were unclear, like in the creative sector, a discussion also emerged about the significance of boundary drawing and why industry-specific manifestations were important: Many people signed the petition, but many have a very vague connection to the industry, but more willingness to be seen and heard, and to be in the limelight that the [...] industry has. Whether someone harasses you in your amateur [context] says more about how society is at large than how […] the industry looks. (Survey answer from organizer 17) The most central aspect of who should be allowed to participate, however, was not about professional identity but about power. Since sexual harassment is seen as an expression of a power structure, where those who have more power take liberties with those with less power, it was seen as important that the participants in the group had a fairly equal status so as not to contribute further to these power schemes. The petition groups' organizers were thus careful not to accept employers or managers as participants, or to mix teachers with students. A closely related dilemma concerns the cases where members were related to persons with power over the others in the group. Sometimes such participants left the group voluntarily as they felt that their presence risked reducing the trust within the group. The importance of other power structures based on age and sexuality was another discussion that came up, as it could sometimes make the requirement for equality within the group complicated. The affinity with other vulnerable people collided with the affinity of colleagues, family, or others of the same age, or of other forms of similarities. I took the initiative to an IRL meeting afterwards; it was very empowering to meet people, but I reacted that most of the people who came to the meeting were heterosexual white women in their 40s and 50s. [. . .] We did not manage to recruit the young, perhaps because one of the members of the admin did not want to have students in [context] because she taught [there]. I thought we should have included the young [. . .] because they are the weakest and perhaps the most vulnerable, at least it has been so historically. (Survey answer from organizer 60) To sum up, although it as was fairly easy to gather around the problem of sexual violence and harassment, there were still conflicts within the movement, where some groups felt that their interpretations of the problem was not sufficiently acknowledged, while others felt excluded altogether. Reconnecting to Spivak's (1988) notion of strategic essentialism, it was undeniably the case that at some level, #metoo activists of different identities agreed on a common problem or shared identity, despite mutual conflicts and contradictions. This was, after all, what made the Swedish #metoo petition groups possible at all. However, to create a community that feels like a "home" and safe space for all concerned, a shared issue is likely not enough.

Clear rules and roles
A strategy that contributed to creating a trusting environment in many petition groups was the development and communication of clear rules.
Clear directives on publishing in the group. We were clear about how to safeguard anonymity and total anonymization of testimonies (no one was allowed/could publish testimonies in the group and testimonies were first sent to e-mail addresses that we admins later published without names and places or other disclosure in the group). Additions to the group needed to be approved by the contact person and us in admin. The group was secret and not searchable. (Survey answer from organizer 5) The rules were communicated to the participants through the active moderation of posted comments. Those who moderated the groups reminded the participants about the rules and the goals of the campaign, thus improving the level of discussion. They also worked actively to ensure that the rules on anonymization and generosity were followed and they closed down discussion threads that did not follow the code of conduct. There were sometimes competing objectives between individuals, where some wanted to share their full stories or warn others about certain perpetrators, while organizers wanted to protect the collective against the consequences of outing names and too many details.
Another safety measure was the establishing of clear communication channels, where one or a few individuals were appointed as spokespeople for the group. The idea behind the measure was on the one hand to control what information was spread about the petition group, but also that the appointment of a spokesperson indicated that this person did not speak or act in their own name, but for the whole group. It was thus a way to remove unwanted attention for individual organizers, by emphasizing the collective voice.
Such rules and strategies were developed and disseminated within the groups as well as between the petition groups, largely via the overall coordination group, which gathered the organizers.

Limitation of information
Another strategy for increasing trust was to anonymize testimonies, so that neither testifiers nor accused perpetrators were possible to identify. Technical affordance was fundamental to effective implementation of petitions. The technical safety was (somewhat surprisingly) not what the organizers experienced as risky; rather, it was the human factor that was sometimes experienced as worrisome: it was crucial that members of the group did not spread names or information.
All groups embraced some degree of anonymity, especially the right to be anonymous, including the right to provide anonymous testimonies. This anonymity mainly concerned what was communicated externally and to other members. Most groups had restricted access to information about the testifiers and to the uncensored testimonies. In the relatively open groups, however, many testimonies were published directly by the testifier without moderation, so the person became known to the whole group, which could consist of thousands of people, sometimes with serious consequences.
To avoid testimonies leaking from the group, we started collecting them in a separate document and deleting them from the Facebook group. This turned out to be too late. A woman was contacted by her perpetrator after her testimony leaked. (Survey answer from organizer 41) Following this event, participants in this group were asked to send their testimonies either directly to the organizers or through a form that allowed full anonymity. Here, different considerations needed to be considered and balanced against each other. While it was important that information did not leak out of the group, the sharing of testimonies and feedback on these stories was important for developing a trusting atmosphere, which encouraged more people to testify. This was resolved in some petition groups by making the administrators share the testimonies on social media, allowing the testifiers to be kept anonymous, while people were still able to discuss the testimony and express their support: We had rigid rules on anonymity in the group, for having the safety to share. This meant that it was mainly us administrators that shared the testimonies in the Facebook group. (Survey answer from organizer 8) The need to remain anonymous seems to have been particularly important in tightly connected networks, where everyone knew everyone, and this for two reasons. First, there was too much to lose if it became known that one had participated in #metoo activism, as there were few possible new workplaces to switch to. Second, the perpetrators' relatives were often well known and included in the targets' social networks, which meant that they (or other people with an interest in the issue such as human resource managers) could easily access information in social media by looking over the shoulders of a partner or simply by sharing login information with a family member.
A few groups maintained full internal anonymityi.e., testimonies were anonymous to the organizers and the organizers were anonymous to the participants. These groups did not use social media for discussions, as this was perceived as too unsafe. Some petition organizers went even further, chosing to keep their petition groups completely anonymous and not naming any organizers or collecting any signatures at all. This approach of total anonymity, even towards journalists and researchers, could create difficulties reaching out and gaining legitimacy, but was sometimes deemed as necessary to avoid reprisals from colleagues and family, or for fear of what the public exposure would entail.

Limitations in scope
Another strategy for creating a safe space applied by several petitions was limitations in scope. The gathering of testimonies and signatures could, for example, be done during a limited time only and the group was then closed down when the petition was published. Another aspect of scope concerns the size of the group. Although it was seen as positive that the petition evoked interest and engagement, problems arose if the group became too large. The speed at which some of the groups grew was difficult to predict and many organizers were taken by surprise: Then we went out with this on Thursday, and started sending out to our contacts, and so. The same evening, we had fifteen hundred members. And in three or four days we had almost thirteen thousand members in the group. (Interview with organizer F) Scaling up a feminist supportive culture in larger groups was challenging. This challenge was made clear by the speed through which groups developed, requiring quick decisions without any formal leadership. The larger the group became, the more unsafe it became for the participants, as the possibilities for information leaks increased. Above all, it was labor intensive to monitor large, sprawling group discussions that went on around the clock. One of the organizers answered our question about safety: 'Our safety was never a problem. The most problematic was workload and stress'. As mentioned by the informant above, groups sometimes grew uncontrollably, leading administrators to feel that they could not administer the group or keep the information from leaking: The larger the group became, the more unsafe it became. We tried to make those who wrote in the group aware of this and think about that we became bigger and that the confidentiality became increasingly difficult to maintain. However, the stories that became public were anonymized and when it [the petition] was published, we closed down the group for reasons of confidentiality. (Survey answer from organizer 19) One measure taken to reduce stress was to "pause" the group for a period when it was not possible to give all participants a much-needed rest from the intense discussions in the forums.

Summary of findings
This study shows how the Swedish #metoo activists developed safe spaces that were distributed over several communication tools, from regular meetings, phone calls, and e-mails, to collaborative writing tools, survey tools, and foremost discussion groups in social media. These safe spaces provided what Fraser (1990) calls subaltern counter-publics where experiences were shared and feminist discourses were developed, formulated, and published in public news media. In conclusion, safe spaces were created through a supportive community and shared values, by separatism, through clear rules and roles, limiting information, and limiting the scope of communication. At an overall level, the groups applied similar strategies for creating a safe space, but there was a varying level of safety that could be divided into three types of groups (Figure 4).
Some groups applied light safety that relied on shared values. Organization and collection of testimonies were made in closed groups on social media to which people belonging to the defined group were invited. The invitations to the group worked according to the snowball principle so that everyone was invited by someone who knew them. That is, there was a social closeness and feeling of having shared values among the participants of these expanded networks, even though the large size of many of the groups meant that most of the participants were strangers to each other.
Other groups had moderate safety where communication was more regulated. On Facebook, the organizers created hidden and "private" groups that were not searchable and that sometimes constantly changed names in an attempt to hide the group. Anyone invited was checked by the group's administrator, who in some cases also moderated posts before they were posted to ensure that no one posted names or anything else that could compromise the safety of individuals. In this strategy for creating a safe space, trust was based on common rules and leadership.
Some groups applied strict safety measures based on alienation where a sometimes completely anonymous organizational group collected testimonies via a survey. Those who participated had no opportunity to contact each other or see the other participants' contributions. Here, the technology was used to minimize the risk that identifying information would spread by minimizing the possibility for the participants to communicate with each other or with the organizers.

Discussion
The political strategy of #metoo was very much about transparency -to make structural discrimination visible by putting words on silenced experiences, often associated with shame and taboo. While the movement was ultimately about creating visibility for these experiences, the form of participation made it enough to write "me too", thereby indicating that one shared experiences and/or felt sympathy for other people's experiences, without the need to define exactly what these experiences Figure 4. Illustration of information flow on three levels of safety: community, where information flow in all directions between the organizers and participants, and participants and participants, and safety is based on trust on shared values and community; regulation, where information flows in both directions between the organizers and participants but where the organizers act as gatekeepers controlling the information flow between participants and participants; alienation, where information flows from participants to organizers, but participants have no means to contact other participants and can be anonymous also for the organizers were. Neither a clear identity, nor testimonies were required to communicate solidarity with the larger collective.
While the safety of participants was important to create trust in the organization of petitions, there was a contradiction between being anonymous and having a sense of community and trust in the collective. Here, the petition groups employed a varying degree of safety that corresponded with the degree of perceived risk from the participants: from a process of trust through shared values, through one based on pronounced regulation and leadership, to a process of trust that involved calculating and applying strict safety measures, where the technology was used to maintain distance and enhance alienation between participants.
None of the organizers was afraid that the technology would fail or lead to any privacy concerns. This tendency is confirmed by previous research: people choose to rely on technology even when they should know better (Guberek et al. 2018). The perceived risks were foremost of a social and cognitive nature, e.g. that information would leak out from the group, or that the information load would become too overwhelming to handle. The organization of the movement can be described as a state of emergency where everything happened quickly and often unplanned, creating great pressure on those involved. As in other crisis situations, trust was not perceived as an issue in relation to the technology, but in relation to people (Light and Miskelly 2019;Tapia and Moore 2014).
One interesting finding was how many petition groups showed a reversed trust process, contradicting previous research where trust is typically seen as a process of crossing distances, and as something that accumulates over time. This process has been described as containing various phases such as developing trust, building trust, and maintaining trust (Corbett and Le Dantec 2018a). In this process, typically each phase is based on the previous one, moving from a situation where there is a lack of community, and trust is in regulations and systems rather than in other people, towards one where trust is based on increased community and belonging, and participants trust each other as they know each other and have developed similar values. In the Swedish #metoo movement, the different types of trust processes employed can be seen as an expression of the levels of trust that existed in the different industries from which the petitions originated. However, it can also be seen as a trust process that went backwards, where trust was high in the beginning when groups were smaller but decreased as the groups became larger and exposed to the public. We could also see how assumptions of community as central for trust was contradicted. The most important risk was actually experienced in groups where the participants came from tightly knit networks where they knew or understood each other well. In these contexts, there was a perceived risk that participants could have conflicting loyalties, as everyone was connected and depended on each other. This shows the importance of acknowledging the inequalities and conflicts within communities to better understand how trust develops in social media applications. It also shows, as previous research also indicates, that generalized trust through shared values and exclusive membership might be more important than social bonding (Moser et al. 2017;Stolle 2002). The discursive process provided by the #metoo movement was here important for creating generalized trust, as global news reports were used in combination with discussions in closed groups to develop and strengthen shared values.
In contrast to the tightly knitted networks, trust was perceived as higher in several of the larger groups, where most people did not know each other, and the network was more loosely connected. Here trust was motivated by a shared belief in the core value of the movement and feminist meeting practice, which reproduced a supportive environment in the online forums and therefore enabled a large-scale trust in relative strangers.
Both the participants and the organizers initiated and participated in processes they had little control over and rarely had previous experiences with. The strength to actually implement these risky projects came from previous successful petitions that acted as role models and established a shared set of feminist practices and examples. Technical affordance was another important factor. The large-scale deliberative action was made possible by the fact that the necessary communication technologies and social networks were in place. A number of easily available technical tools functioned as prostheses that enabled the organization to be scaled up to thousands of participants. Several petitions were made in horizontal networks in social media that organized women in industry-or interest groups. In many cases, the basis for the petition groups were found in already existing networks that enabled the groups to be formed and spread quickly and easily. These loose networks of people, brought together by technology and shared interests, enabled people to find a way to meet and focus on the one thing they had in common. The Swedish #metoo petitions thus followed a political model focusing on shared difference (Mohanty 2003), and by doing so, it succeeded in temporarily uniting people who were sometimes in conflict in other matters.

Conclusion
In this article, we have explored the rapid processes through which the Swedish #metoo petitions were created, where over hundred thousand of participants coordinated themselves under a common manifesto to share personal and often traumatic experiences with relative strangers. Our results confirm previous research, showing how digital feminist activism can be risky, strenuous, and overwhelming (Mendes et al. 2018). To overcome such obstacles, it was central for the organization of the petitions to create safe spaces, and various strategies were used to promote trust and ensure safety.
According to Haraway (1991), technology can be seen as a kind of prosthesis, which, if we trust it, allows us to reach what was previously unattainable. One central finding of our analysis was that the trust that enabled so many to participate in the Swedish #metoo movement and break the culture of silence around sexual harassment and abuse, was not trust in a particular technology. Rather, the trust was aggregated, consisting of subsets of trust distributed over time and space. Trust was established in different technical systems, institutions, people, shared values and practices, as well as a large number of trustgenerating interactions over time, both before and during the actual organization of the petitions. Another central finding concerns how trust was created. Our results show that trust decreased over time as the movement grew and public exposure increased, a trend that was most evident when participants came from a tightly knit community, where the risk was being exposed within the community.
Contrary to simplistic beliefs of community as being necessary to build trust, this study suggests that a certain amount of alienation in the interfaces can actually be significant in order to organize around important societal issues. In our studied case, the need for transparency and community was balanced with the need for anonymity and distance.
We conclude by acknowledging some of the limitations of the study. First, the #metoo movement's expression and development differ considerably between national contexts. The study at hand is limited to Sweden and is thus not representative for the movement at large. Secondly, while one strength of qualitative studies such as this one, is that results are grounded in informants' experiences of activism rather than their use of a particular technical platform, it is also a limitation, as we may have missed important design details, such as how specific interfaces affected the participants' micro-actions. We therefore hope to see more research examining these issues from a design perspective, that seek to explore the possibilities to develop safe yet dynamic spaces for large-scale deliberation and activism, especially those that involve potential risks and conflicts.  Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creativecommons.org/licenses/by/4.0/.