Introduction

One of the criticisms of VR is that it can be a rather solitary experience. Users are cut off from the physical world, seeing and hearing different things from those around them. Most of the projects we have discussed thus far have focused on a single user experience. We turn now to reflect on the research potential offered by social VR systems, which allow multiple users, normally not in the same physical location, to come together in a shared virtual environment.

Writing about potential future applications of VR in the early 1990s, Valerie Stone (1993), argued that collaborative and creative experiences would be of much greater value than the kind of solitary and non-creative interactions that she associated with the video games of the time. The European Commission-funded COVEN (Collaborative Virtual ENvironments) project, which ran from 1995 to 1999, was a pioneering attempt to create a networked VR infrastructure to produce these kinds of group experiences, with projects such as Slater et al (2000) examining the potential social effects of these shared virtual spaces. The dream of VR becoming a collaborative platform has, however, only really become possible since the third wave of VR from the early 2010s, thanks to dramatically improved technology, commercial interest and a much larger consumer audience. Indeed, Perry (2016) optimistically described social VR as the ‘killer app’ that would allow the new generation of VR to break into the mainstream.

Online collaborative social experiences in virtual environments have been around for some time, though they were at first available only via desktop interfaces. In the early 2000s, VR developer Bernie Roehl (2019) was involved as a beta tester on the development of Second Life (Linden Lab, 2003), the best-known and most successful of these desktop-based collaborative environments. He recalls how difficult it was to explain to friends at the time that, while Second Life looked somewhat like a game, it did not function as such because there were no particular goals to meet. Instead, platforms like Second Life act as spaces for socialising, allowing synchronous engagement with real people from around the world meeting in virtual locations. Newer platforms, such as AltspaceVR (Microsoft, 2015), Hubs (Mozilla, 2018) and Facebook Horizon (Facebook, 2019) take this model one step further by allowing users to immerse themselves in these social worlds via an HMD.

Roehl (2019: 291) identifies four key qualities underpinning social VR platforms: multiple users are present and represented by avatars; users can navigate the virtual environment; users can communicate with each other; and users are able to tell which of the avatars is communicating with them. With desktop platforms such as Second Life and its VR successors, however, there are significant technical constraints that shape their design and interactive qualities. Users’ avatars are generally graphically simple because of issues around latency when the software is trying to draw multiple people interacting in the same space. Nonetheless, these constraints create opportunities for creating new ways to interact socially. Mcveigh-Shultz and Ibister (2021) even go so far as to describe these limitations as creating new forms of ‘weird social’, allowing for experimentation with new modes of social interaction. In AltspaceVR, for example, you can see multiple emojis floating above a user’s avatar to express their feelings in a given moment.

In this chapter, we explore the different ways in which collaborative virtual experiences can be used within research projects. In some cases, this involves researchers building custom software for their experiments – something we will return to in Chapter 6. For those lacking this kind of technical expertise, however, there are many opportunities to employ commercial social VR platforms within projects, many of which are free to use. These commercial platforms benefit from being well designed and robust, with an established user base, and having some capacity for customising environments and avatars and opportunities for interaction around the needs of a research project. Beyond these advantages, as Maloney et al (2021) point out, the large numbers of commercial social VR users interact with these platforms in ways that their designers never anticipated – sleeping in VR for example. As such, the communities and modes of socialisation enabled by commercial social VR platforms become fascinating objects of study.

The chapter starts by exploring the opportunities for collaboration between participants and researchers presented by social VR. Next, we explore the role of the avatar in shaping interactions between social VR users, looking in particular at questions around social cues and harassment. Finally, we report on a case study from our own work examining the communities that have grown up around a virtual church within AltspaceVR.

Opportunities for collaboration

One of the traditional selling points of VR was a vision of people in different geographic locations being able to come together in the same virtual space for education, work and leisure. The reality certainly is not as seamless as the vision; social VR is yet to become the ‘killer app’ that Perry predicted. Nonetheless, the steady stream of work on social VR became a flood after 2020 as the COVID-19 pandemic trapped people in their homes, and tools for remote collaboration were suddenly in high demand. Rzeszewski and Evans (2020) undertook a review of how attitudes toward VRChat had changed since the beginning of the pandemic. Methodologically, this is an interesting use of content analysis, using the Steamworks application programming interface to scrape reviews of the software posted on Steam – the most widely used platform for purchasing and downloading PC games. The 28,334 reviews downloaded were split into pre- and post-March 2020 and subject to qualitative analysis using NVivo. The reviews posted after March 2020 contained fewer gripes about the glitchiness of VRChat and an increased sense of positivity toward the platform and its potential. This included reflections on how engaging with VRChat helped some users overcome wider social anxieties by creating a safe online space to socialise during a period when meeting friends in the physical world had become actively dangerous.

At the time of writing, there are a number of competing social VR platforms, which have a different appeal to different audiences; Rec Room (2016), for example, is targeted more toward children. Common features include opportunities to play games collaboratively, interact with objects and chat. These platforms have different limitations on interaction, which are useful to consider when designing a project employing one or more of them. Liu (2020) has produced a useful comparative overview, asking participants to reflect on the difficulties of attempting the same set of tasks in different platforms. Common issues included not being able to communicate with another user if they were not in the same virtual room – which made coordinating activities quite difficult. Indeed, in our own experiments with AltspaceVR we have found that getting people to understand how to navigate around and communicate with other users can be quite tricky, although the platform does have a useful onboarding process to train novices.

The uses to which people put these platforms varies considerably and there is clearly scope for projects that simply undertake ethnographic work about how social VR shapes forms of socialisation. Zamanifard and Freeman (2019), for example, searched for social media posts about using social VR in long-distance relationships, examining how users were able to create a sense of intimacy through the illusion of co-presence. There remain, however, real barriers to reproducing the conversational flow experienced when physically in the same space. Bleakley et al (2020), for example, highlight the need to develop technical systems that more consistently create a feeling of social presence, such as being able to understand the role a given user is playing in a social interaction. They also identify practical issues that act as social barriers, such as not being able to see what other users are looking at in a virtual environment, thus reducing the illusion of sharing an experience.

We will return to the question of shared social cues when talking about avatar design. It is important to note, however, that the limitations of the platforms in reproducing non-verbal communication have a significant impact on how we are able to research topics around VR collaboration. Nonetheless, there has been some interesting work in recruiting existing social VR users as research participants. This has the advantage that they are already familiar with VR and have their own equipment and so can participate remotely – something which was very useful during the COVID-19 pandemic. Saffo et al (2021) recruited participants this way for a replication study employing bespoke VR environments that participants could access from home. That the researchers were able to replicate results from earlier work using both quantitative and qualitative approaches is quite encouraging. One possible reason why this worked effectively is that the recruits were familiar with VR and thus were not distracted by the technology when undertaking the study. Indeed, ethnographic work by McVeigh-Schultz et al (2018) highlights how social VR users have to learn a new set of social cues in order to successfully interact with those spaces. Again, this is significant when thinking about how to recruit participants into VR studies, particularly if placing novices into virtual settings with unfamiliar social protocols.

Industrial interest in VR products has often been in its use for training purposes, where shared and collaborative experiences can add real value. Employees can enter dangerous or potentially expensive scenarios and gain experience before undertaking these for real. Such training was initially intended for an individual working alone, but, as the technology is maturing, so more collaborative scenarios are starting to be produced. The ways in which industrial users are applying collaborative VR is an important topic of study in itself, which can be explored through a combination of discussions with designers and participants as well as ethnographic observations. In a review of industrial VR applications, Berg and Vance (2017) highlight the example of the firm Case New Holland, which brought together its engineering, design and marketing teams within a VR environment in order to work together on optimising new products. Indeed, this idea of collaborative design within VR is itself an interesting research technique to explore.

A fascinating, if somewhat quirky example of how co-design can be undertaken in VR is Mei et al’s (2021) CakeVR tool. Here, the researchers prototyped a system to solve a common problem experienced by pastry chefs, where clients struggle to articulate precisely how they want their customised cake to look and, as a result, are not always happy with the final outcome. The researchers undertook a storyboarding exercise with pastry chefs and used this as the starting point for a basic VR platform that allowed chef and client to come together to build models of potential cakes and visualise them in the round. They then tested this in VR with a researcher role-playing with chefs and clients to explore the useability and usefulness of the system; each participant spent about an hour in VR co-designing a cake. Although only a pilot, with limited graphical sophistication and flexibility, both groups of participants were enthusiastic about the potential for collaborative cake design using this medium. Given the emphasis in a great deal of social sciences and humanities work on the importance of co-design with participants (Zamenopoulos and Alexiou, 2018), one can see a similar approach being taken in work with, say, residents and housing developers or creative practitioners and community groups. Although CakeVR was a custom-built tool, there is potential for a simpler approach where an existing social VR platform is used to undertake different kinds of co-design activities. Many of the commercial platforms allow varying degrees of creative work where users can come together to build and interact with shared objects and environments, meaning that there is real potential for rapidly prototyping ideas without the need for investing in custom software.

Collaborative and sharing experiences are central to some forms of educational pedagogy (Le et al, 2018). Undertaking this kind of work in VR can sometimes demand that a complex and expensive custom environment be constructed. The kind of ‘flying classroom’ described by Schulte et al (2018), for example, needs significant investment because of the way in which it meets highly specialised training needs. Here, they developed a system whereby a remote instructor could adopt the same point of view as a trainee and appear as a set of virtual hands, guiding the trainee through the correct actions to repair a piece of industrial machinery. This kind of approach can be incredibly valuable, particularly in allowing experts with highly specialised knowledge to be sent to remote sites. There are some limitations to it, however, beyond the simple cost of developing a custom environment for the specific training needs. One of the issues is where very accurate movements need to be replicated by the trainee. Xue et al (2020), for example, have piloted a system where patients with arthritis can be brought into a virtual surgical room to talk with a remote doctor about the process of injecting themselves as part of a treatment regime. Correctly using a hypodermic needle, however, requires fine motor skills, and a purely visual VR simulation would be less effective in this use case. Xue et al therefore combined their collaborative VR platform with a SenseGlove, a device that mechanically manipulates the movement of the fingers, looking a little like a nightmarish spider clamped around the hand. This gives haptic feedback to the user such that it gives the sensation of picking up a physical object. As a result, patients in this study were actually able to get a feel for giving themselves injections, with a remote medical professional being able to watch and give them feedback in real time.

Tools like the SenseGlove can seem quite magical, but are expensive, require custom software to be built for each application, and their appearance can be intimidating for the faint-hearted participant. Indeed, for many applications, the SenseGlove would be overkill, though there has been some really interesting work looking at object handling in a virtual museum using the system to simulate the feel of holding a piece of pottery (Senseglove, 2021).

Other researchers examining the educational potential of social VR take the more cost-effective approach of using existing platforms. Yoshimura and Borst (2021), for example, taught a course in Mozilla Hubs and then asked their students for feedback on how well it worked. They emphasise that those students who did not get simulator sickness found the sense of presence particularly useful, especially in the context of the COVID-19 pandemic prohibiting conventional face-to-face teaching. Students also reported feeling less nervous giving presentations as avatars than they would when using conventional video. Interestingly, the students noted the lack of visual cues from the teacher’s avatar had an impact on their feeling of being properly engaged with. We will return to the issue of avatar design later.

Returning to Stone’s (1993) original idea that VR should be both collaborative and creative, there have been some innovative projects attempting to engage with more arts-led approaches. Gochfeld et al’s (2018) interactive theatre piece Holojam in Wonderland is a good example of this. They ask what added value can be brought to theatre by using VR, beyond the merely gimmicky. Drawing on the Alice in Wonderland stories, they built a virtual environment in which a live audience was present in the same virtual space as real actors controlling avatars in a short play. Using VR allows the actors and audience to change in size, reflecting the shrinking and growing tropes of the original stories, with Gochfeld et al reflecting that VR offers productive opportunities to set stories in magical and fantasy worlds. This is a nice example of where a bespoke virtual environment is more than just a necessary expense to test a particular hypothesis, and is itself a creative output underpinning the wider theatrical experience.

Engaging participants’ own creative energies can also be quite productive. Baker et al (2021) have undertaken a large project examining how social VR can be used with older people. They used a participatory action research approach, working with older people to discuss the kinds of materials that they would like to experience within VR. The researchers then built a prototype environment, based on participants’ interest in creating a tool for reminiscing. Their School days app allowed participants living in different parts of Victoria, Australia, to come together within a virtual classroom, to reminisce about their youth and to share experiences. By adopting a participatory approach, the team were able to explore how this kind of tool could be refined and made more useful for helping older people socialise and avoid becoming isolated even when living some distance apart.

Avatars, social cues and harassment

The School days app was developed as part of the Ageing and Avatars study funded by the Australian Research Council. Considerable effort went into testing different ways to reproduce participants’ bodily movements within social VR in part to try to resolve some of the issues around ambiguous social cues and non-verbal communication highlighted earlier. Some of their participants felt, however, that highly accurate body tracking might actually reproduce negative stereotypes of ageing in the virtual world, such as uncontrollable trembling (Baker et al, 2019). Again, this emphasises how participatory design work can be incredibly valuable in highlighting issues that might simply have never occurred to young, fit programmers.

Another issue raised by the Ageing and Avatars project was that participants found the avatars’ lack of facial expressions somewhat disconcerting. This is a question that has generated considerable research interest, including a large project funded by the European Commission, VR Together.¹ One of the challenges that the VR Together project tackled was trying to find ways of creating more photorealistic avatars in VR (Gunkel et al, 2018). Participants were asked to watch a video together within a VR environment, simulating the effect of sitting with other people sharing a movie (Simone et al, 2019). The participants found that the experience was better where the avatar sitting next to them was a photorealistic representation created using Microsoft’s Kinect sensor, which tracks body movement. It has to be said that looking at the project’s promotional videos, the effect is somewhat rough around the edges, producing a low-resolution rendering of users wearing HMDs appearing as avatars in the virtual scene. Nonetheless, the project has examined some significant questions, such as how the lack of facial expressions on avatars acts as a barrier for more naturalistic social interaction, including examples of discussing shared photographs (Li et al, 2019), or understanding what another avatar is looking at in a virtual environment (Rothe et al, 2020).

While it may appear at first to be a somewhat niche concern, the way that participants are represented within a social VR environment can be quite significant for how projects can be undertaken. Indeed, the fact that most social VR platforms use rather cartoonish or abstract avatars has been seen as a barrier to business use, given the desire to project a more serious and professional appearance (Lee et al, 2021). Nonetheless, even with these somewhat abstract designs, avatars can create a reasonable sense of being socially present with other people. Yoon et al (2019) tested six different avatar types from live video and whole-body capture through to cartoonish virtual characters. They found that participants responded best to a whole-body avatar, though abstract representations of just the upper body or a cartoon style could also be effective in different types of collaborative activities. While it is now easier than ever to create photorealistic models of individual participants using technologies such as volumetric capture and photogrammetry, these remain fairly expensive and complex to operationalise. It is therefore reassuring to know that relying on the basic cartoon avatars of commercial social VR products can still be effective when researching virtual collaboration.

The presence of even a somewhat abstract avatar can be seen to create a much greater sense of collaboration within a virtual space. Herder et al (2019), for example, set up a task where participants had to operate virtual machinery. Participants found it easier to complete the task when the researcher appeared alongside them as an avatar compared with when they were merely talked through the exercise by a disembodied voice. The avatars used in this study were basic rather than fully photorealistic models, again giving reassurance that a sense of co-presence and collaboration can be generated in research studies with abstract representations of other users. This does not get away from the fact, however, that most non-verbal social cues are missing from these types of systems. There are potential technical fixes for this – Izzouzi and Steed (2021), for example, suggest that HMDs with built-in eye tracking could be used to reproduce a user’s gaze through animating eyes of avatars, while prototypes have been built showing how a combination of cameras pointing at the mouth with artificial intelligence reconstruction could be used to create VR avatars that have more natural facial expressions (Olszewski et al, 2016). These kinds of technical solutions are just beginning to be seen in some commercial products. As a result, questions around non-verbal communication within social VR pose an interesting and rapidly evolving set of challenges to researchers. Maloney et al (2020b) highlight how the use of an HMD and the associated sense of physical presence give many more opportunities for creating non-verbal cues than in desktop platforms such as Second Life. Nodding, bodily position and the use of emojis are central to signposting emotional engagement with other social VR users, and there is further potential for improved hand and finger tracking to capture gestures.

Methodologically, Divine Maloney’s work has relied on a mix of interviews and participant observation within social VR to explore the everyday use of these shared virtual spaces. One of the important issues that his work highlights is how users of social VR reveal and conceal personal information in order to protect their personal safety. Children in particular are at risk of intrusive questioning and even grooming within these spaces (Maloney et al, 2020a). This work highlights the significant ethical concerns around working with participants in social VR spaces, particularly for projects relying on ethnography. There have been quite a few studies looking at how social VR is used by vulnerable or marginalised groups to socialise and gain support from like-minded people. Acena and Freeman (2021), for example, have looked at the ways that LGBTQ users have built communities within social VR, undertaking a series of interviews and observations of online events. The ethical risks of this kind of work need careful consideration, particularly given that some participants may be living in real-world communities where their sexual and gender identities are seen as problematic or even illegal.

While some marginalised communities have attempted to carve out safe spaces within social VR, it is important to note that harassment and trolling are significant problems on these platforms. At its most benign, this might be a simple case of new users not really understanding the implicit rules of social interaction in an unfamiliar platform. This being said, in a study by Blackwell et al (2019), most participants who experienced harassment in social VR distinguished between a naïve violation of established mores and malicious intent. Within their interviews, participants recounted the kind of racist and sexist abuse that is familiar to anyone who has studied wider online and gaming cultures (Mantilla, 2013) but is no less depressing for its ubiquity. What VR brings to online harassment, however, is a sense of embodied proximity. Abusive users placing their avatar far too close to others, for example, can create real discomfort for those targeted (Sun et al, 2021). Teleporting into a virtual space to spew invective at people is also common: these types of abuses come without the risk of being subject to the kind of physical altercation that such behaviour might trigger in the real world.

Thus, when one is considering using a commercially available social VR platform for research, there is a risk of coming into contact with abusive activity. Developers are not unaware of this as a problem and many platforms include tools that allow users to mute abusers and even make them disappear from view. Indeed, some have specifically made choices around the avatar design options available to users, for example avoiding hyper-sexualised attributes in order to reduce sexist harassment (Kolesnichenko et al, 2019). When working with participants in social VR, therefore, there may have to be compromises in order to maximise participant safety, depending on the needs of the research project. Undertaking data collection within a social VR space brings the advantage of better understanding how participants interact with that environment, but it can bring the risk of trolling from random users unrelated to the project. Activities outside that space, such as interviews on Zoom, can risk the anonymity of participants from vulnerable groups, but allows for more in-depth conversations with lower risk of hostile interruption.

Case study: VR Church

We turn now to reflect on one of our projects, led by Natasha, undertaking an ethnographic investigation of the communities formed around church services that take place in VR. Religious groups have a long history of building communities in digital worlds, including Second Life (Radde-Antweiler, 2008) and on social media (Cheong, 2014). Yet VR potentially offers a more immersive religious experience and community. Founded in 2016 by DJ Soto, VR Church offers a new and innovative opportunity where people can use VR to develop real relationships across geographical boundaries (Round, 2019; Jun, 2020). The VR Church community is an extremely active group with weekly services in ‘Church World’ via AltspaceVR and various chat platforms including Discord.

Natasha attended European and US services in AltspaceVR using the now discontinued Oculus Go HMD. This was a stand-alone headset with a single hand controller and three degrees of freedom, meaning that it tracked head orientation but not body movement. The project took the form of an autoethnography, with the aim of exploring the practices and experiences of worship within social VR. This was supported by semi-structured interviews with members of the congregation, and content analysis of the group’s Discord server. This combination of methods enabled a reflection upon the researcher’s experiences as part of the VR community and the perspectives of the congregation.

The autoethnography was undertaken at weekly services between November 2018 and June 2019. Natasha entered the church community openly as a researcher and constructed her AltSpaceVR avatar to personify her identity as a young, white, middle-class, cis woman (Figure 4.1), but chose not to reveal her religious positionality in a deliberate effort to avoid religious debates. Some members of the congregation chose avatars that attempted to match their physical appearance, while others preferred to construct their own visual identities, thus enabling a sense of anonymity in the virtual space (Galanxhi and Nah, 2007). Despite this ability to customise one’s appearance, the HMD used by an individual shaped the capabilities of their avatar. Natasha was using the Oculus Go with a single controller, meaning that her avatar only had one hand, whereas those using more expensive devices with two controllers had two hands: “Like right now I’m using the Go which is obvious. I’m trying to upgrade to the Quest when it comes out, which is gonna be like $400, if I’m correct, so I’ll have two hands” (Interview with David in VR).

Natasha’s avatar in a VR Church service. Note the single floating hand giving away her use of a cheaper headset. This image was produced prior to a significant graphical upgrade to the platform in 2020

Source: Natasha Keen and AltspaceVR (2020)

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Natasha’s avatar in a VR Church service. Note the single floating hand giving away her use of a cheaper headset. This image was produced prior to a significant graphical upgrade to the platform in 2020

Source: Natasha Keen and AltspaceVR (2020)

• Download Figure
• Download figure as PowerPoint slide

Natasha’s avatar in a VR Church service. Note the single floating hand giving away her use of a cheaper headset. This image was produced prior to a significant graphical upgrade to the platform in 2020

Source: Natasha Keen and AltspaceVR (2020)

• Download Figure
• Download figure as PowerPoint slide

In this example, Natasha was unable to hear the music alongside the video feed, so felt that she was missing out on a fun social experience. Regular members of the congregation were used to these technical glitches, but they clearly served as a barrier to engagement.

Another key problem that worshippers discussed was the impact of trolls coming into the service and yelling or making a nuisance of themselves. As a relatively high-profile event within AltspaceVR, these interruptions were not uncommon, meaning that the organisers had become well versed in muting and ejecting those trolls. The existence of trolling was significant particularly because some members of the group were active in VR Church precisely because they had issues around anxiety or other vulnerabilities that made it more difficult to access church spaces in the real world. The role of social VR as a safe space for vulnerable individuals again emphasises the importance of giving very careful ethical consideration to how an ethnography is undertaken with these groups.

Conclusions

The prevalence of glitchiness and trolling are just some of the factors that mean social VR remains an imperfect realisation of the promise to enable people to come together in a virtual space for creative socialisation. Nonetheless, what VR Church and other communities within social VR demonstrate is that VR cannot be considered a purely solitary experience. Many of the communities within social VR platforms are public and open to visitors. This creates a wealth of opportunities for data collection via ethnographic work within the platforms themselves and online interviewing, as well as content analysis of the chat logs and social media posts generated by these communities.

Because of technical constraints around processing power and bandwidth, the avatars used in most of the commercial products are rather abstract or cartoonish. They generally lack subtle ways to convey non-verbal communication and other social cues, which can act as a barrier to engagement. Nonetheless, various experiments have demonstrated that, while more realistic models do generate a greater sense of embodied presence, even abstract or cartoony avatars can effectively produce feelings of co-presence and a capacity for socialisation among users. This means researchers can be relatively confident that even the commercial social VR platforms offer good opportunities for exploring new forms of collaboration and socialisation within these virtual worlds.

Significant research projects have been undertaken where custom environments have been built that can be shared between two or more users. As we will discuss in Chapter 6, these offer excellent opportunities to examine participant response to specific scenarios when considering interactions with other people. The need to invest in creating a custom environment will, of course, depend on the demands of the research project and available budget. What we wish to emphasise here is that commercial social VR platforms can be a good vehicle with which to test ideas – and indeed to work with large numbers of potential participants – before going to the expense of developing custom software. Moreover, the huge expansion of interest in social VR sparked by the COVID-19 pandemic will inevitably drive further innovation in this area, meaning that collaboration within VR will remain a significant and rapidly evolving research topic for many years to come.