Proxemics and Social Interactions in an Instrumented Virtual Reality Workshop

1 INTRODUCTION

• Virtual environments may support social activities as remote work and pressure to reduce travel increase, but are not replacements for face-to-face interaction.Source: §1 · p. 2
• Virtual environments differ in affordances, fidelity, scale, and accessibility, complicating organizers’ choice of tools for their intended experience.The paper prioritizes small-group interaction, networking, and open-ended discussion for its academic workshop.Source: §1 · p. 2
• The study combines position and orientation logging, observations, and interviews to examine social interaction across keynote, breakout, and networking activities.It also releases code and a captured dataset for collection and analysis.Source: §1 · pp. 2–3
• The authors ask whether physical-world proxemics help analyze virtual interaction, how groups relate to activity and environment, and which social cues help or hinder workshops.Source: §1 · p. 2
• The workshop study uses an open-source Mozilla Hubs environment customized to log position and movement while supporting keynote, breakout, and informal networking activities.Source: §1 · p. 2

2.1 Methods for Analysing Behaviour in Virtual Environments

• Prior VE behavior research relies mainly on observations, questionnaires, or communication logs; this study combines positional instrumentation with observations and interviews for a small scientific event.Source: §2.1 · p. 3

2.2 Proximity and Personal Space

• Proxemics links interpersonal distance to nonverbal communication, while personal-space patterns also vary with factors such as context, relationships, and gaze.Hall’s model names intimate, personal, social, and public distance zones, but the cited work notes that virtual and physical-space findings are not fully consistent.Source: §2.2 · p. 4

2.3 Forming Groups and Social Activities

• Environment design affects group formation: large open spaces can lack the social density that makes contact easy, while cohesion and social identity are also recognized influences.Group formation can involve initiation, identifying members, and negotiating group preferences and goals.Source: §2.3 · p. 5
• Virtual environments can create new interaction forms, while studies report that some social norms, including interpersonal distance and gaze, also appear in virtual settings.Source: §2.3 · p. 5
• Avatar realism, distinctiveness, gaze, and limited social cues shape social VR experiences, and users may compensate through other behaviors such as exaggerated movement.Source: §2.3 · p. 6

3 RESEARCH QUESTIONS AND VIRTUAL WORKSHOP DESIGN

• The study asks whether physical-world proxemics apply in virtual settings, how groups relate to activity and environment, and which cues or artefacts help or hinder workshop experience.Source: §3 · p. 6

3.1 On-Boarding Sessions

• On-boarding let participants test equipment, learn Mozilla Hubs, and practice movement including flying, with a warm-up session for last-minute technical issues.Source: §3.1 · p. 6

3.2 Workshop Programme

• The programme included a keynote, participant pitches, small-group breakouts, presentations, and generous breaks intended for rest and informal social interaction.Source: §3.2 · p. 6

3.3 Workshop Environments

• The large Outdoor Meetup space hosted keynotes and informal breaks, while three linked Lake Office rooms supported small-group discussions.Outdoor Meetup measured 70 by 40 meters; Lake Office rooms measured 20 by 30 meters.Source: Figure 2 · p. 7, §3.3 · p. 7

3.4 Informed Consent

• Participants consented to tracking, observations, photographs, and video, but chat and audio were not recorded except for the livestreamed keynote; interviews were optional.Source: §3.4 · p. 7

4 INSTRUMENTATION, DATA, AND INTERVIEWS

• A modified Mozilla Hubs client logged positional data, alongside workshop observations and post-workshop semi-structured interviews; code, dataset, and analysis scripts are openly available.Source: §4 · p. 7

4.1 Instrumented Virtual Environment

• Client-side logging captured player identifiers, timestamps, positions, orientations, frame rate, mute and speaking status, and other interaction signals on browser ticks.The system batched ticks before POSTing data to avoid socket failures from frequent per-tick requests.Source: §4.1 · p. 8

4.2 Proxemic Dataset

• After resampling, the dataset contained 37,457 position logs across 6,619 frames from 26 users and four virtual spaces.Resampling standardized data to ten frames per minute because participants’ frame rates varied.Source: §4.2 · pp. 8–9

4.4 Semi-structured Interviews

• For each frame, the researchers calculated pairwise 3D Euclidean distance and angle using position and direction data; nine participants completed 45-minute interviews.Source: §4.2 · p. 9, §4.4 · p. 9

4.3 Observations

• The 4.5-hour workshop activities ran as planned, and 14–18 participants often remained in the virtual space during breaks to network.Some early frozen-frame and audio problems were resolved by refreshing and rejoining; no breakout-room technical problems were reported.Source: §4.3 · p. 9

5.1 Social Cues and Proximity in Virtual Space

• Breakout participants formed cohesive small groups at personal or social distances, while keynote participants were less cohesive and extended into public space.Intimate-distance collisions were rare in both sessions, and participants moved away when they occurred.Source: §5.1 · p. 10, Figure 4 · p. 10
• Participants found attention and turn-taking harder to read without facial and eye cues; overlapping audio and uncertainty about private chat also complicated conversations.HMD head and hand tracking provided additional social signals, and participants described breakout-room conversations as easier.Source: §5.1 · pp. 10–11

5.2 Groups and Social Interactions

• Workshop activity concentrated in the Outdoor Meetup centre; participants avoided floor decals and the amphitheatre, which hosted no planned activity.The authors note that the amphitheatre’s limited interactivity and resources were consistent with prior research.Source: §5.2 · p. 11, Figure 5 · p. 11
• 54% of keynote time, participants kept the speaker in view; sixteen attendees occupied 20×15 meters of the 70×40-meter room and used fly-mode for personal spacing.Breakout rooms’ low ceilings accompanied closer spacing and limited flying, although flying through the ceiling was possible.Source: Figure 6 · p. 12, Figure 7 · p. 12, §5.2 · p. 12
• At the first break, participants formed one large group; by the second, smaller groups surrounded a large group, and participants more often stayed on the floor plane.As participants became familiar with the controls and sounds, they formed smaller groups and used avatar positioning or orientation to attract or signal attention.Source: Figure 8 · p. 13, §5.2 · p. 13
• Avatar networking made it harder to associate avatars with people and encounter them by chance, while some participants felt avatars reduced anxiety about approaching senior colleagues.Some participants wanted realistic, expressive avatars to help with networking and remembering others; others valued feeling less identifiable or judged.Source: §5.2 · pp. 13–14

5.3 Play and Finding Familiarity

• Participants used virtual-ocean play and selfies to connect, but Mozilla Hubs’ novelty also brought navigation and feature-use difficulties.Onboarding helped some participants learn controls and social mechanisms, though participants were not always sure how to behave or manage audio.Source: §5.3 · p. 14

5.4 Limitations

• The live Social VR workshop drew many domain-interested participants, and technical issues persisted despite onboarding; browser and HMD users were not compared in controlled conditions.The authors note that participant interest shaped some qualitative findings and that future work could use controlled experiments and audio signals.Source: §5.4 · pp. 14–15

• Participants received onboarding positively, and the authors recommend a short written introduction and a pre-workshop session to familiarise attendees with the platform and social norms.For some participants, unfamiliarity with Mozilla Hubs may have helped establish new workshop norms.Source: §6 · p. 15
• Movement between linked rooms lacked visibility of participants’ destinations, and the authors call for feedback that makes destination state and movement more visible.A participant entering a portal could appear to others to have disconnected; increased social translucence is proposed to reduce this disruption.Source: §6 · pp. 15–16
• The authors suggest that awareness of who is attending and where, alongside cues for turn-taking and joining groups, could help people initiate conversations.They also note that incidental features such as the tiki corner and ocean appeared to encourage ice-breaking interactions.Source: §6 · p. 16
• Virtual-space use could diverge from design intent: the keynote screen and podium went unused, while the floor texture influenced participant positioning.Participants avoided the floor texture and appeared to move along its edge; the authors note that dynamic reconfiguration can produce unanticipated occupancy and group formations.Source: §6 · p. 16
• Personal space varied with activity and environment: flying appeared regularly during the keynote, but not in breakouts, and was rare during informal breaks.The authors suggest perceived room containment or social expectations may discourage flying outside the keynote.Source: §6 · p. 16
• HMD users could gesture more expressively, while desktop users could share materials and take notes more easily, producing different workshop experiences across devices.Moderators were assigned as guides regardless of device, and the authors identify a need for customisation or intervention to minimise imbalanced experiences.Source: §6 · p. 17

• The workshop findings suggest proxemic interactions in virtual environments can be congruent with physical interactions, while flying can add a dimension to personal space.Group formations adapted to room size, and the authors report that textures, shapes, and routing can affect movement.Source: §7 · p. 17
• The authors identify a need for expression, indicators, and social translucence when using a generic virtual environment such as Mozilla Hubs.They also note that open-source instrumentation can provide position and action data for studying social behaviour and exploring generative layouts.Source: §7 · p. 17
• The authors conclude that remote events can benefit from virtual environments, within the scope of their academic-workshop findings.Source: §7 · p. 17

• The supplemental materials include JavaScript code for instrumenting Mozilla Hubs Cloud for data logging.Two detailed instrumented items, 4.1(viii) and 4.1(ix), were later additions and were unavailable during the study.Source: A SUPPLEMENTAL MATERIALS · p. 20
• The materials include the anonymous dataset collected from the workshop.Source: A SUPPLEMENTAL MATERIALS · p. 20
• The materials include Python Jupyter notebooks for cleaning and analysing data and generating figures and an animation; the authors provide an ACM Digital Library and GitHub access location.Source: A SUPPLEMENTAL MATERIALS · p. 20