Learning as the Audience Grows

This post provides details on the social features we developed prior to our launch on Steam Early Access and the work we’ve targeted based on our learnings since then.

High Fidelity development is broken down into ten priority areas. Ahead of adding High Fidelity to Steam Early Access in November 2016, we focused heavily on our “Community Permissions and Protections” feature area. This project encompasses things like the provisioning of temporary rez and edit rights for domain visitors but, more importantly for our launch, also includes features that help our users protect themselves from abuse and harassment.

Based on prior review of the current literature around virtual reality spaces, we developed a “personal space bubble” that we anticipated would be needed as the audience in High Fidelity domains increased.

Almost by its nature, griefing behavior involves testing the limits of a system. The personal space bubble has already been iterated extensively to make its function clearer to new users, provide visible cues for users that it is present, and to handle cases where griefing users have experimented with disrupting others by interacting with the bubble in unexpected ways. A short demo of it below:

As our audience increased after launch on Steam, our initial findings were that approximately one in twenty visitors to our Welcome area would engage in mild to extreme disruptive behavior. In some cases, this behavior is the result of adjusting to being in VR, or being unaccustomed to interacting with others in close quarters. In these cases, and in cases where visitors are especially loud, sometimes due to inexperience with their mics, we often found that users would adjust their behavior quickly when asked to.

We also found that a few of the disruptive users were actually intent on griefing others as their primary activity. We had anticipated this based on our research on other VR applications on Steam.

Our initial findings after launch were that:

• Even with only a few other users present, actively responding to griefing users (through muting or ignoring) was difficult, in part because users would need to be ready for unexpected behavior continuously.
• Tracking and kicking griefing users based on tracking the location of their avatars is more difficult in practice than we anticipated. Griefing users quickly became adept at changing their avatar size, griefing from a distance, and in some cases creating invisible avatars

Since launching on Steam, High Fidelity has begun hosting events of increasing scale, including a recent event with over 100 people in the same virtual space. In environments with interactions on this scale, additional protection and controls are required. We’ve focused in two areas:

• Deploying the People Action List (PAL) that allows users to manage the interaction with others at scale. You can see a simplified demonstration of the PAL below

The PAL interface is designed to accommodate interaction with a large group of other users. We continue to iterate on PAL to make it easier to identify and respond to disruptive users within a crowd.

• Adding additional custom controls to domains to help users manage specific experiences. While exciting, giving users the ability to edit and move objects in a VR domain is in some cases an invitation to act mischievously. To be successful, the platform will require relatively fine grained control on the size and type of actions that visiting users can do. We’ve added the ability for domain owners to set limits on the kinds of behaviors allowed in their domains and within zones within each domain.

The Future

As the community continues to grow, we anticipate that we will also need to support “opt-in” communities of users that will provide the equivalent to a “reputation” that others can use to predict user behavior. We hope features like this will form the basis of a kind of “shared trust” between established users.

We’ll continue to update the community as we learn more.