VR Ecosystem and User Scenarios

A better analogy: the arena

Returning to this paper years later, the ecosystem analogy in Challenge 1 still does one job well and another badly, and it took me a while to see why. Ecology is excellent at describing what kinds of things exist in a world — substrate, producers, consumers. That is why the abiotic/biotic split and data-as-producer hold up. But ecology has no vocabulary for how much each agent may change the world. Trophic levels rank who eats whom, not who may rearrange the furniture. I was asking one analogy to do two jobs, and the seam shows at the bottom rung: I leaned on microbes to represent immobility, when motility is precisely the thing many microbes have.

A shared space with a single event in it works far better for the second job. Picture a concert in an arena, with the stage in the centre rather than at one end.

The show — music and light — is the shared content. It is spatial, multimodal and ambient: it fills the volume and reaches everyone at once, though differently depending on where they stand. The performers are the only agents who can alter it. The floor audience moves wherever it likes and changes nothing. The stands hold an audience with their own view of the same show, fixed to the seat they chose from a finite set. All four are present simultaneously, by the architecture of the building rather than by a rule — which is what the farm could never manage, since penned and free-ranging animals are not in the same field as the farmer.

And the arena covers something the original paper treated separately. A concert is also broadcast, and the viewer at home is not a fourth role — it is the second flavour of Passive Observer this paper already named: controller POV or static camera. The person in the stands is the static camera, holding their own fixed viewpoint. The person at home is the controller POV, watching a viewpoint someone else selected.

Laying it out this way exposes something I had left implicit. The ladder is not one variable with three settings; it is two separate increments:

fixed position → free position → free position plus authority over shared state

Stands to floor is purely positional. Floor to stage is authority. That distinction matters in a real system, because the two permissions are separately grantable: you can let someone roam without letting them touch anything, which is the entire reason the middle role exists.

It also produces a genuinely counterintuitive result. The viewer at home often sees more than the person in the stands — close-ups, camera changes, replays, the mix straight off the desk. Lowest permission, highest information. Anyone designing an observer experience should not assume that more freedom yields a better one; an active observer who has wandered behind the data may understand it less well than a passive observer following a well-chosen view.

A fourth axis: the return channel

There is one more thing the arena separates that the original paper never did, and it only becomes visible once the broadcast audience is in the picture. It is not what you can perceive. It is whether you can be perceived.

The floor crowd is felt: a band demonstrably plays differently to a live room. The stands are still seen and heard — they cannot move, but forty thousand people roaring is unmistakably present to the performer. The remote viewer is perceptually absent. Same show, same content, and from the stage they do not exist. In current terms, they are attending with camera and microphone off.

This has a name. Benford and Fahlén's spatial model of interaction splits presence into aura, focus and nimbus — nimbus being how much of you is projected to others, how observable you are. Awareness between two participants is a function of one's focus and the other's nimbus. That is exactly the pair of properties the arena pulls apart, and the vocabulary is older than this paper by two decades; I arrived at the distinction independently, but I did not invent it.

One precision matters here. The floor's influence is not authority. They cannot change a note. What they change is the person who changes the notes. Influence through the return channel is mediated by the controller; authority is direct. That is why crowd energy does not promote the floor to Controller — and it is the cleanest available answer to why the middle role is genuinely a middle role rather than a weak Controller.

The completed framework

Put together, the properties that were tangled through the original paper separate into two groups. Three describe a participant; two describe the session they are in. Keeping them apart is what the earlier version failed to do — Table 2 mixed roles with what were really session-level properties.

The three roles this paper named in 2017 are a diagonal through the participant axes, not a scale. Controller is full authority, free vantage, high nimbus. Active Observer is no authority, free vantage, visible. Passive Observer is no authority, fixed or assigned vantage, and nimbus that varies from present to absent depending on whether they are in the room or watching a stream. Once the axes are separate, combinations the roles never covered become expressible — and most of the interesting ones are exactly the combinations nobody has designed for.

Where each piece came from

These axes were not invented here; they accumulated. Setting them against their sources shows which questions kept recurring:

The chart this became at Microsoft

Four years after the Spatial OS paper, consulting for Microsoft Research on hybrid meetings, I built a matrix of every configuration a hybrid meeting could take. It had three axes — Embodiment (levels of agency based on representation), Spaces (levels of control based on ownership), and Interfaces (levels of immersion) — with each of the 48 resulting nodes coloured by how feasible it was to build.

I did not notice at the time that I had reached for my own vocabulary. Spaces is the ownership ladder from the AR Cloud layer of the 2020 paper, relabelled from personal/private/public to Personal/Communal/Virtual. Embodiment is the axis this 2026 note calls nimbus — how much of you is projected to the people you are with. And Interfaces is the display-options question from 2017, generalised: not screen or headset, but a graded scale of immersion.

The original was a lattice you walked around in VR. ShapesXR deleted the model when their billing changed and I was not on a paid plan; what survives is a screen capture, from which the axes were recovered. Rebuilt here as three panels, because on a page you cannot walk around a cube.

Two things fall out of it that were not visible in the lattice. The Avatar column is feasible almost everywhere — across every interface and every kind of space. The Physical column is difficult everywhere except face-to-face, which is close to a tautology once stated but was not obvious before: being bodily present is not something an interface can grant you. Between those two, the whole design space collapses onto a single question — if you cannot be there, how well can you be represented?

That is the nimbus axis, discovered a second time and under a different name.

Two places the analogy would mislead

At a concert the controller authors the content live, whereas in a shared virtual environment the dataset generally pre-exists and the controller acts upon it. What carries across is not where the content comes from, but who may alter what everyone else is experiencing.

More importantly: at a concert, the remote viewer's absence is physics. In a virtual environment it is a setting. The return channel exists, and somebody chose to leave it off by default. That single substitution — a design decision wearing the costume of a physical constraint — is, I think, the reason hybrid meetings go wrong. The asymmetry looks like a property of the medium, so nobody revisits it.

Open questions

Who holds authority over the view? Someone chooses what the home audience sees, and it is not the band. The vision mixer has authority over presentation without any authority over content: they cannot change a note, but they decide what a million people look at. None of the three roles account for that. It sits sideways to the ladder — less than a Controller over the data, more than an Observer over the framing. In a shared session that role is real: the facilitator who repositions everyone's view, the guide who directs attention, whoever defines the fixed camera positions a passive observer chooses between.

Is nimbus grantable per participant, or only per session? Broadcast bundles "many receivers" with "no return path," but those are separable, and unbundling them is where equitable hybrid participation would have to start.

This paper identified one missing middle role in 2017. Nine years later the arena surfaces two more — the presentation authority, and the participant whose presence is switched off by a default nobody chose deliberately. The recurring problem, across all of it, is the middle.