Visualizing Placement Proposals for Window Arrangement in Mixed Reality: A Comparative User Study

Imagine you are trying to organize a messy room, but instead of a floor and walls, you are floating in a giant, invisible 3D space. This is the world of Mixed Reality (MR), where computers project digital windows, menus, and tools right into your view, hovering in mid-air. It sounds like magic, but it comes with a tricky problem: where do you put all these floating windows? If you just let a computer decide for you, it might put them in spots that are hard to reach or block your view of the real world. If you try to move them all yourself with your hands, your arms might get tired, and you might spend too much time just arranging things instead of doing your work. This paper dives into the "Goldilocks zone" of this problem: finding a way to let the computer suggest good spots for your windows, but still letting you pick the one you like best. It's like having a helpful friend who arranges your books on a shelf and says, "Here are four great spots for this new book," rather than just shoving the book somewhere or making you climb a ladder to find a spot yourself.

The researchers from the University of Konstanz wanted to see which way of showing these "suggested spots" works best. They tested four different methods in a virtual reality game where participants had to plan a trip. The goal was to arrange seven different windows (for flights, hotels, and schedules) as quickly and comfortably as possible.

The first method was the Manual Positioning baseline. This is the "do-it-yourself" approach. When a new window popped up, the user had to grab it with their hand and drag it to wherever they wanted. It's like moving furniture around a room with your bare hands—total freedom, but it takes time and effort.

The other three methods were the "suggestion" techniques, where the computer offered four specific spots to choose from, but showed them in different ways:

1. Situated Icon Preview: The computer placed a tiny, floating cube with an icon on it at each of the four suggested spots. It was like seeing four small sticky notes on the wall telling you, "You could put the window here." It was quick to spot but didn't show you what the window would actually look like.
2. Situated Window Preview: This was similar, but instead of a tiny icon, the computer showed a full-size, empty frame of the window at each spot. It was like seeing four ghostly outlines of the furniture already placed in the room, so you could tell if it would fit or block a lamp.
3. 3D Preview: This was the most unique one. Instead of looking at the spots in the room, the user looked at a tiny, miniature model of the whole room (a "world-in-miniature") floating in their hand. The suggested spots were marked on this tiny map. It was like looking at a dollhouse version of your room to see where the furniture fits best before placing the real thing.

The study found some surprising results. When it came to speed, the suggestion methods were much faster than doing it all manually. The "Situated Window Preview" (the ghostly outlines) was the fastest of all, taking about 54 seconds to arrange the windows, while the "Manual Positioning" took over 120 seconds—more than double the time! The "3D Preview" was slower than the other suggestions, taking about 79 seconds, likely because users had to get used to looking at the tiny map.

However, here is the twist: even though the suggestions were faster, the participants overwhelmingly preferred the Manual Positioning method. When asked which one they liked best, they gave the "do-it-yourself" approach the highest rating (an average of 4.46 out of 5), while the suggestion methods scored much lower.

Why would people choose the slower, more tiring option? The researchers found that people felt they had more control when they could drag the windows themselves. They also felt that the suggestions sometimes didn't give them enough information about why a spot was chosen, or they just weren't used to the new way of doing things. It turns out that even though the computer's suggestions were efficient, users didn't trust them enough to give up their own control. The study suggests that the best solution might be a "hybrid" approach: let the computer offer a few great starting spots to save time, but make it super easy for the user to tweak and move them exactly where they want, combining the speed of automation with the comfort of human control.