XR headsets increasingly host ordinary 2D user interfaces — windows floating in a physical room. Their usability then depends on conditions the designer does not control: how far away the user is standing, what the lighting is doing, and what physical object is in the way.
AdapTest, from Jungwoo Rhee, Guhn Lee, Yumin Jung, Mustafa Doga Dogan, Justin Moon and Andrea Bianchi, is an in-situ authoring and testing tool for exactly that problem, accepted to PACM HCI ISS 2026 (the ACM DOI, 10.1145/3859325, is listed on the preprint but not yet resolving).
The gap it names
The paper’s diagnosis is one sentence and it is the useful part: “Current design workflows often separate how a variant is designed from how it actually looks in the target environment.”
That is the standard pipeline. Lay out the interface on a desktop, build it, put the headset on, look at it, take the headset off, adjust, repeat. Each loop crosses a context boundary, and every crossing loses information — you cannot see the room while editing, and you cannot edit while seeing the room.
The consequence is not that designs come out ugly. It is that environmental adaptation gets designed by assumption. You imagine a lighting condition, pick a contrast value for it, and find out later.
The three variables
Drawing on prior research and a formative study with 12 interface designers, the system targets three environmental variables:
- Distance — how far the viewer is from the panel
- Lighting — ambient brightness and its effect on legibility
- Occlusion — physical objects between viewer and interface
These are the right three, and the third is the one desktop tools handle worst. Distance and lighting have rough analogues in conventional responsive design — viewport size, dark mode. Occlusion has none. There is no web equivalent of “a chair is in front of the lower third of your interface,” and no way to anticipate it from a monitor.
What the tool does
AdapTest lets designers select interface elements, define when and how they change, compare the resulting variations, and test their activation — without coding.
The “without coding” part is load-bearing. Environmentally responsive behaviour is normally expressed as conditional logic against sensor values, which puts it on the engineering side of a team and out of reach of the person making the design decisions. A direct-manipulation tool moves the decision back to the designer, in the room where the condition occurs.
“Compare the resulting variations” matters too. The hard question in adaptive design is not what a single variant looks like but whether the set of variants is coherent — whether the interface reads as one thing across conditions, and whether two conditions firing at once produce something sensible.
The evaluation, and the honest result
AdapTest was tested against a desktop design plus in-headset preview workflow in a within-subjects study with 12 designers — the realistic baseline rather than a straw man.
Compared with the baseline, designers:
- Accounted for a wider range of environmental factors
- Addressed more overlapping conditions
- Repeatedly tested and revised their designs in place
And the finding the authors report plainly: there were no significant differences in workload or usability despite longer sessions.
That last clause is the one to notice. Sessions ran longer, and measured workload did not rise. Longer and not harder is a specific and credible shape for a result like this — the designers were doing more work because more of the problem had become visible and addressable, not because the tool was fighting them. A paper claiming the tool was both more thorough and faster would be harder to believe.
The “more overlapping conditions” result is the substantive one. Dim light and partial occlusion and three metres away is where adaptive interfaces actually fail, and it is precisely the combination nobody models at a desk, because you have to be standing in it to notice it exists.
What to take from it if you build XR work
Most readers here are not building productivity panels, but the mechanism generalises to anything with legible content in a headset — an annotation layer over a sculpture, a wayfinding overlay, a score for a performer.
Author in the space. If your work will be seen in a particular room, the design decisions about contrast, scale and placement belong in that room. This paper is evidence that doing so changes what designers attend to, not merely how comfortable they are.
Occlusion is a first-class condition. Gallery visitors stand in front of things. Other visitors are physical objects. If your overlay assumes an unobstructed view, it assumes an empty room.
Test the combinations, not the conditions. Each variable alone is tractable. The failures live in the overlaps.
It is worth noting the limits: 12 designers, one tool, one task domain, and a comparison against one particular baseline workflow. What the study establishes is that in-situ authoring changes designer behaviour in measurable ways — not how much better the resulting interfaces are for end users, which is a different experiment.
Sources
- AdapTest: Supporting Designers in Authoring and Evaluating Environmentally Responsive User Interfaces for Extended Reality — arXiv:2610.21386
- arXiv DOI: 10.48550/arXiv.2610.21386
- Andrea Bianchi — Make Lab, KAIST