Each has a different viewpoint, so reference must be made explicit

What it is

Saying “the one on the left” or “this” over a real table often works, because both people see the same furniture. Put headsets on, and left is the speaker’s left; the other person’s field of view may not contain that virtual object at all, or their own panel may hide it. Spatial deixis in co-located mixed reality can no longer lean on a default shared view. “Which one” has to become a mark the other person can also see.

Coordinates already aligned only guarantee the object sits at one physical place. They do not guarantee that the sentence lands on the same object.

Why it happens

Deictic words are egocentric: this / that / left / right bind to the speaker’s body and gaze, not to the room. In real collaboration the other person can see the hand, the facing, the thing being stared at, and can translate the egocentric phrase into an object in the room. Virtual objects break that tacit deal: they may exist only inside a headset, the other field of view is narrower, and UI may hide the same object. “This” becomes a gamble.

Making reference explicit is dragging the referent off the speaker and onto the object: highlight that instance, give it a short name, draw a world line from the speaker to it. The highlight must be world-locked. A mark glued to the other person’s head does not help — a turn takes the mark off the object.

Studying it

A collaborative identification task: a guide refers to several similar-looking virtual objects in speech, a follower selects. Conditions: speech only (“the one on the left”), plus a world-locked highlight, plus an object name. CSCW work on referring (Fussell and colleagues on shared visual space) takes misunderstanding counts and clarification turns as the core dependent measures.

Independent variables: referring channel (speech / highlight / name), whether the two face the same way, whether the objects look alike. Dependent variables: wrong selections, clarification turns, time to finish.

When two people stand shoulder to shoulder, the cost of egocentric speech drops. A face-to-face stance must be a separate cell, or the problem is underestimated.

Where it stops holding

Objects with unique names (“hydrant,” “number 7”) already make speech enough; highlight is redundant. Demos where two people stay side by side with highly overlapping views can still get away with left/right now and then. Remote voice plus a shared 2D map moves the referring problem onto the map legend; it is no longer a headset viewpoint gap. If the product’s own copy keeps saying “look to the left,” it ships the same class of error even after human collaborators have learned to stop.

Applying it

• In collaboration, do not let “left / right / this” be the only referring channel. On select, put a world-locked highlight on the object so it is visible in the other person’s view.
• Give discussed objects a short, stable name or number that speech can repeat.
• Change system prompts to the object name or a highlight. Do not write “left side of the screen.”
• How to check: stand two people face to face; one says “that one,” the other taps. Wrong selections that happen on speech alone should drop once a highlight is added. If they do not, the highlight was glued to the speaker’s view, not to the object.

Related

• Same group：N5.05.1 Several people need a shared coordinate reference · N5.05.3 Permissions on shared content must be clear
• Nearby：N5.10 Shared Space and Multi-user Collaboration · N3.08 World-locked, Body-locked, and Head-locked
• Search terms：spatial deixis · referential ambiguity · world-locked highlight