Social VR Safety: New Research Insights into Risks and Protections in New Online Spaces

Are We Safe in Virtual Reality? A recent study has delved into Social Virtual Reality platforms like VRChat, RecRoom, HorizonWorlds, and AltSpaceVR. Researchers identified emerging safety risks and evaluated existing protective mechanisms by analyzing 212 YouTube videos in hopes of creating a more secure and inclusive virtual social environment in the future. Dive into the full paper here.

Examples of Identified Safety Risks: a) Virtual Violence: The attacker’s virtual actions cause real-life physical fall for the victim. b) Virtual Abuse: Profanity and body-shaming directed at a snacking victim. c) Virtual Sexual Harassment: Deceptive virtual objects used to harass the victim.

What Are the Existing Safety Risks in Social VR?

In this work, researchers analyzed 212 YouTube videos and their transcripts that depict real-life social VR interactions. This in-depth look revealed a broad spectrum of safety risks, from harassment to identity theft.

An Overview: Definitions of Safety Risks in Social VR

Researchers found that social VR users face unique safety challenges shaped by new social norms and cultural values. Here’s a quick rundown:

1. Role-playing: This causes confusion between reality and make-believe, sometimes leading to fraudulent impersonation.
2. Immersive Dwellers: Users who loiter or sleep in public VR spaces can become targets of abusive behaviors from others.
3. Misread Cues: Miscommunication in VR can lead to misunderstandings, such as mistaking friendliness for unwanted social interactions.
4. Misused Safety Features: Intended safeguards like vote-kick features can be exploited for malicious actions.
5. Minors Picking on Adults: The blend of adults and minors in VR can lead to a surprising trend where minors often become perpetrators.

Some other safety risks are similar to those that have occurred in other online contexts. Still, the immersive experiences and enhanced avatar control in social VR can amplify their severity and cause greater harm. See more examples below.

Examples of Safety Risks: a) Virtual Voice Trolling: A female user switches to a masculine voice, startling a male friend. b) Virtual Scaring: The attacker uses a frightening avatar to chase and intimidate the victim. c) Virtual Trash Actions: The victim uses a “cross” gesture unsuccessfully to ward off an intrusive attacker.

How do Victims, Attackers, Bystanders, and Spectators React to These Safety Risks in Social VR?

An added complexity to these findings is the varied user reactions, from different groups ranging from empathy to endorsement of risky actions, which underscores that safety in social VR is shaped by a wider social context beyond just the immediate user.

In this work, researchers thus analyzed reactions from four types of users to safety risks in social VR:

• Attackers: These are the users who initiate harmful actions in virtual settings in social VR.
• Victims: These individuals are on the receiving end, suffering harm due to the unsafe behaviors initiated by attackers in social VR.
• Bystanders: This is a critical but often overlooked role. These users are present during the attack but neither initiate harm nor suffer from it.
• Spectators: This role is not directly involved in the VR experience. Instead, these individuals watch the incidents unfold on non-social VR platforms, such as YouTube videos and share their views by posting comments.

The findings are organized below. The middle column categorizes the risk—be it verbal abuse, violence, or harassment—and counts the number of relevant videos documenting such risks. The columns on either side map out a detailed set of social cues, one focusing on attackers and victims.

Social Cues of Attackers and Victims in Each Type of Risks (Identified in Youtube videos posted 2022/01/01–2022/06/01).

Key Takeaways for Social VR Platforms from This Work:

1. Attackers use multi-modal social cues, customized avatar effects, and virtual objects to make the most threatening attacks.
2. Victims instinctively try to shield themselves as if facing a real-world threat, only to discover such defenses don’t work in VR.
3. Most bystanders either ignore the situation or exit the scene. Some laugh, while a few exert social pressure on attackers, indicating a complex bystander dynamic.
4. YouTube viewers debate the ethical implications, especially regarding minors. They also express a range of emotional reactions, including distress, after watching these incidents.

Are Existing Safety Features Sufficient for People to Manage Safety Risks in Social VR?

The second part of the study turns to existing safety mechanisms in social VR. After evaluating a wide range of safety features—blocking, reporting, and more—the study categorized three primary types of safety features:

1. Social Boundaries Settings: For proactive user comfort
2. Quick Reactions: For immediate response to safety violations
3. User Agreements: For outlining community norms and potential risks

High-level Analysis of Safety Features Across Social VR Platforms

Researchers then mapped these existing safety features with the safety risks in social VR they identified in this work. Their findings show that current safety features focus too much on physical proximity and lack natural, intuitive gestures for quick user response, calling for a shift from reactive to preventative safety features:

Mapping Safety Risks with Safety Features

A Call for Reimagined Safety Protocols

Are current safety features adequate? The research suggests they’re not and calls for a rethinking of safety design. The need of the hour isn’t merely adding more features but considering the unique interaction dynamics:

1. Social VR’s immersive quality can make simulated physical risks feel like actual bodily harm.
2. The extent of personal information shared through avatar customization can open the door for more targeted attacks.
3. Poorly designed features can lead to counterintuitive safety risks that defy common sense.
4. Safety concerns aren’t solely due to design flaws but are also influenced by varying community standards and the absence of guidelines in ambiguous scenarios.
5. User perceptions of social norms in VR, which can deviate from offline norms, further complicate safety risk management.
6. The potential for abuse or misuse of existing safety features presents additional challenges to maintaining a safe environment.

Looking Forward: Proactive Safety Measures

Researchers also underscored the need for user-friendly, proactive safety measures on evolving social VR platforms.

NPCs: From Educators to Conflict Mediators

NPCs as Safety Companions. Based on the real video scenarios, the research suggests Non-Player Characters (NPCs) as educators or mediators. In personal learning modes, NPCs can teach users how to make risky objects transparent, while in public spaces, they can intervene in brewing conflicts.

Gesture-Activated Safety

Intuitive Gesture-Activated Safety. Based on the real video scenarios, the paper also proposes more intuitive mechanisms, like covering one’s ears to activate mute settings or using a push-away gesture to maintain personal space.

As social VR becomes increasingly integrated into our online social connections, especially for the younger generation, ensuring user safety in these spaces isn’t just a technological challenge; it’s a societal imperative. This research serves as a checkpoint for more discussions on how we can create truly inclusive and safe VR environments.

If you’re interested in learning more or collaborating in this space, feel free to reach out to me (Qingxiao Zheng) and my amazing co-authors 🙌🏻— Shengyang Xu, Lingqing Wang, Yiliu Tang, Rohan Salvi, Dr. Guo Freeman, and Dr. Yun Huang.