1. Techniques for Interacting with Off-Screen Content
Pourang Irani Carl Gutwin
University of Manitoba University of Saskatchewan
Grant Partridge Mahtab Nezhadasl
University of Manitoba University of Manitoba
Good afternoon.
My name is Pourang Irani. I am presenting work I have done in collaboration with Carl Gutwin and our students Grant Partridge and Mahtab Nezhadasl.

2. Introduction R M R M R R R M R M R
- It is very common to work with documents and workspaces that are larger than the user’s viewport. When we work with text documents, drawing applications or even simply viewing large images, we work with workspaces that span several viewports. With the rapid introduction of devices such as cell phone and PDAs, this is becoming very common.
As a result a significant amount of information that is needed resides off-screen
For instance, a common scenario could be that your friend asks you to meet her at a restaurant near the north gate metro station. To find this location the user has to start exploring various regions of the workspace. If the region is outside the viewing region then a certain amount of navigation may be necessary until the required information is located.
In general, the issue relates to the idea of wanting access to some information, somewhere out of the viewable region or out of sight.
Although techniques such as zooming and panning exist for browsing off-screen content, very little is known how they perform on a range of tasks
We also want to know whether any one technique really works best for a range of different tasks R R R M R M R

3. 2D navigation: Time Multiplexing
Zooming R R M
Time multiplexing techniques allow users to interact with different regions of the off-screen space at different times, in other words off-screen space is inspected in a serial fashion.
There are a number of time-multiplexing techniques which include: scrolling, panning and zooming
In the case of zooming, the user needs to move back and forth between different scales of the workspace in order to find the content of interest
One potentially major drawback with time-multiplexing techniques is that \the user has to perform a significant amount of navigation to locate items of interest.

4. 2D navigation: Space Multiplexing
Overview+Detail DragMag R M R M
Space multi-plexing techniques allow users to concurrently view different regions of the workspace at the expense of screen real estate. These include: O+D, Focus+Context and Portal based techniques.
With techniques such as Overview + detail or DragMag there is the additional complexity in managing the overview and the detail windows.

5. Proxy-based techniques
In these techniques, users interact with objects in the workspace through an intermediary proxy
Proxy based techniques were introduced for inspecting remote objects or objects that are placed far form the user’s workspace
The main ideas behind proxies is that they bring replicas of distant objects closer to the user’s interaction space, for further inspection
This allows the user to see details of the target before performing the task, either targeting, or dropping
In the case of drag-and-pop the user has to perform small hand movements to connect objects that are related, either for dropping objects onto others or for picking remote objects
Vacuum is another technique designed by Bezerianos and Balakrishnan, that also relies on the idea of proxies for selecting remote content.
[Bezerianos and Balakrishnan, 2005]
[Baudisch et al., 2003]

6. 2D navigation: Proxy Hop (Halo + Proxies) M R R R
Hop is a proxy-based technique that allows the user to navigate to off-screen regions by first inspecting the targets. In Hop you have halos, with which you can interact to create proxies and then use the proxies to navigate off-screen. M R

7. Tasks Baudisch & Rosenholtz [2003] Spatially Absolute
“Position”
“Closest”
“Avoid”
Spatially Absolute
Existence, Count, Location
Spatially Relative
Proximity to reference, Proximity b/w objects, Cluster
We were interested in testing these various techniques not with only one task, but with a series of tasks.
We began by categorizing these tasks.
We started off with a set of tasks suggest by Baudisch & Rosenholtz, which they collected based on interviews with PDA users employing maps. These were “position”, “closest”, “avoid”
We extended their set of tasks, into two abstract categories:
Spatially absolute
These describe the relationship between objects and the workspace
Spatially relative
These involve some form of relationship between different objects, and not b/w the objects and the workspace
There are certainly many possible ways to categorize tasks, and many other tasks, such as traversing, route planning, etc. However we selected these as a starting point.

8. Evaluation: Conditions
Navigation Techniques
Zoom - two-level zoom
DragMag
Hop
Tasks
Absolute
Relative
We selected techniques to represent the different classes of systems. Zoom for time-multiplexing, DragMag for space-multiplexing, and Hop for proxy based.
The experiment was performed on a 800x600 resolution with a TabletPC, so a pen-based device.

9. Zooming

10. DragMag

11. Hop

12. Results: Completion Time
We found that users were twice as fast with Hop as DragMag or Zoom, on spatially absolute tasks.
Participants also performed constantly the same with Hop on absolute tasks.
However we noticed that participants performed better with Zoom than Hop or DragMag on spatially relative tasks.
What is particularly interesting is that performance with DragMag always fell between these two extreme techniques on both types of tasks.

13. Possible explanations – strategy
Number of operations about 1/5 with hop
Zooming requires more “trips” …
not so good for absolute tasks
… but good for spatial information
useful for relative tasks
DragMag reduced the number of “trips”
allows users to perform relative tasks with ease
but added complexity of managing windows
Hop was good with absolute tasks because it significantly reduction the amount of navigation necessary. This was primarily a result of using proxies.
Zooming on the other hand required more “trips” but this was helpful with relative tasks but not so good for simpler absolute tasks. Primarily a result of different scales that does not disorient users.
DragMag reduced the number of trips we observed with zoom but the added complexity made it cumbersome for absolute tasks. This was primarily a result of using different scales simultaneously.

14. Possible explanations – task-based
Relative tasks
Need orientation and comparisons b/w views
Rely on VSTM
Absolute tasks
Require information about the objects
Do not require spatial/orientation information

15. WinHop (hop + windows)
We wanted to maintain the favourable properties of these different techniques and design hybrid systems. This led to the design of WinHop and MultiscaleZoom.

16. WinHop

17. WinHop

18. WinHop

19. WinHop

20. WinHop

21. WinHop (video)

22. MultiscaleZoom

23. MultiscaleZoom
Take advantage of properties from Time-Multiplexing (zoom) and Proxy based

24. MultiscaleZoom (video)

25. Absolute: Location (winhop)

26. Relative: Cluster (msz)

27. Results
MultiScale zooms outperforms the other techniques on both types of tasks. Interestingly performance is pretty much uniform with multi-scale zoom.
Winhop comes in second, but primarily for the spatially absolute tasks. We see that for the Location task, Hop performs better than WinHop. This is primarily due to the fact that in hop users were teleporting but in WinHop they were using the window or portal.
In Spatially relative tasks WinHop did not perform any better than Zoom, but significantly better than Hop.

28. Limitations of Hybrid Techniques
WinHop
Many operations, significant learning curve
MultiScale Zoom
Clutter from proxies

29. Main Findings Each task requires a different navigation strategy
Proxy-based ideal for target-only info tasks
Time/Space multiplexing ideal for target-target info
Hybrids improve performance in both types of tasks
Performance with Multiscale Zoom remained constant

30. Conclusion Beneficial to investigate techniques on multiple tasks
Ideal technique gives target and context information
Hybrid techniques are reliably good
Consider multiscale zoom for small displays
Hybrids performance on other tasks