Spatial Approach

2.2.2.1 Taskbar/Dock

The Taskbar (Figure 2.5 and 2.6) or dock (Figure 2.7) follows a spatial approach in its organization of window buttons (each containing an icon and a truncated piece of the window title). Users can access any open window by clicking on a button or iconic representation of the window. The location of icons or buttons on the Taskbar is fixed and therefore this approach takes advantage of the user’s spatial memory. In addition, the Taskbar can group by application when more than a certain number of windows belonging to one particular application are opened. In this case, they collapse into a single application button that activates a pop-up with a list of windows. Recent research has hinted that theTaskbar has potential usability problems Hutchings & Stasko(2003).

One issue raised by some participants in a qualitative study is that when eight or more windows are open, and theTaskbar is in its default position at the bottom of the monitor, only a few (or none!) of the letters in the windows’titles are visible (In the default case, once the number of windows reaches a certain amount (10 − 15 windows for a 1024 x 768 pixel resolution display), very little, if any, of the title bar can be read and only the

2.2 Window Switching Techniques

Figure 2.4: RelAltTab includes two types of background coloring: salmon for the semanti- cally related windows and light-blue for the temporally related windows. It also displays a red number on top of each semantically related window. This number allows the user to directly switch to the window by pressing Alt + NUMBER.

icons remain). In that case, it would need more time to distinguish the windows which have the same icons when using the Taskbar for window switching (visual similarity is increased). Many researchers, including Mircosoft themselves, have attempted to improve the Taskbar. Both Smith et al. Smith et al. (2003) and Robertson et al. Robertson et al.(2004) claim that the grouping by application behavior confuses many users because application windows may not be related to the same task. Smith et al. presented the GroupBar Smithet al.(2003) as a solution, which allows user to arbitrarily assign groups.

Visual TaskTip (Figure 2.8) was developed to enhance the cognition of buttons on the Taskbar, when the mouse is hovered over Taskbar, it displays a thumbnail preview of the opened window¹. However there was no report of any form evaluation of the tool.

Figure 2.5: Ungrouped window buttons on the Taskbar

1http://www.pctipsbox.com/visual-task-tip-for-windows-xp/

Figure 2.6: Grouped application buttons on the Taskbar

Figure 2.7: Dock tool in Mac OS, which is used to launch applications, and switch between running applications.

2.2.2.2 Expos´e

Expos´e ¹ is another example of a spatial approach, which tiles all opened windows or those of the focused application so that they are all visible at once Apple. But as the number of opened windows increases, the legibility of their content decreases, making them hard to distinguish. Expos´e uses a non-stable spatial layout to arrange the opened windows in a layout: when a window is repositioned or resized between two invocations of the technique, the layout may completely change. When users want to switch back to a window previously switched, they may no longer find it at its last location, wasting the benefit of spatial memory.

2.2.2.3 EyeExpos´e

The EyeExpos´e system Kumar et al.(2007) presented an innovative extension of the Expos´e system by allowing users to use a combination of keyboard and eye gaze. This approach combines the use of a two-dimensional layout visualization for showing to the user all opened applications and the use of eye gaze tracking for selecting the window of interest. The authors designed an experiment to evaluate EyeExpos´e and found Alt+Tab in Windows XP that was faster when the number of open windows was low (4), and EyeExpos´e had the lowest switching time when the number of open windows was high (12). ForAlt+Tab, its performance scaled worse relative to the other methods evaluated as the number of open windows increased. However we were not able to find other research

1http://www.apple.com/pro/tips/switch expose.html

2.2 Window Switching Techniques

Figure 2.8: Visual TaskTip, it displays a thumbnail preview of the opened windows when the mouse is hovered over Taskbar.

Figure 2.9: Expos´e view of open applications.

evaluating the relative performance ofAlt+Tabto support their findings. We could notice that they did not take into account the visual similarity of windows in the experiment, which directly impacts the visual search time for finding the target window.

Figure 2.10: TheTaskpos´e visualization arranges open windows in two dimensions when the visualization is called up. Windows automatically size relative to their importance, and closely-related windows appear together.

2.2.2.4 Taskpos´e

Taskpos´e Bernstein et al. (2008b) (Figure 2.10) is another example of a spatial ap- proach, which is a screen-filling visualization of user’s workspace in two-dimensions, rep- resenting opened windows by thumbnails and using a degree-of-relatedness to implicitly create task groupings. It uses the WindowRank algorithm and window switch history to define the window association, and then using this association heuristic to layout related windows near each other. It uses a non-stable spatial layout to arrange the opened win- dows and the size of the thumbnails can also be changed. When users want to revisit a window, the size and its last location may have changed, wasting the benefit of spatial memory. However when users work on multiple tasks simultaneously, Taskpos´e would move the tasks close together to make them difficult to distinguish. Meanwhile it does not adequately use the space to display thumbnails as big as they can (the big thumbnail can reduce the selection time). During longitudinal evaluation, authors observed that Taskpos´e was most useful when the number of opened windows outstripped the space available on the WindowsTaskbar, In that case, the performance of Taskbar will greatly decrease (see Section 2.2.2.1).

2.2 Window Switching Techniques

2.2.2.5 SCOTZ

Keith Humm presented Spatially Consistent Thumbnails Zones (SCOTZ, see Figure 2.11) which is an example of a spatial approach and uses the stable zones to display window thumbnails Humm (2007). SCOTZ divides the entire screen space into a grid of equally- sized zones. Each zone represents a group of thumbnails based on the same application, and each thumbnail only belongs to one zone. Especially, zones are always located in the same space and never move. Taken into account that windows are opened and closed frequently, spacing inside zones must be much more dynamic than the zones themselves.

The author use a hybrid approach between zone sizing and space-filling techniques: a zone is sized based on the nearest square number to the number of items required. When resize occurs, the current grid is transposed to the upper left coordinates of the new grid.

The author evaluated SCOTZ and found that the performance of SCOTZ performs faster than the other techniques (Grouped Taskbar (see Figure 2.6 ), Ungrouped Taskbar (see Figure2.5 ), andAlt+Tab (see Figure 2.1 )) for both low (4, 8) and high window density (16, 32). SCOTZ is more similar to Expos´e than Taskbar and Alt+Tab (SCOTZ has group mechanism and the property of spatial constancy, and Expos´e may have bigger size thumbnails), however the author did not compare with theExpos´e, leading to results not convincing. Meanwhile the author did not take into account the visual similarity of windows in the experiment. This factor can also impact the results of the experiment.

Taket al. added thesize morphing (see Figure2.12) operation to theSCOTZ (see Fig- ure2.11) technique, thesize morphingoperation allows for the addition of new items while maintaining as much spatial stability as possible and allocating more space to frequently- used applications and windows in order to reduce their selection time Tak et al.(2009a).

The authors investigated the performance impact of four different orderings (spatially stable, recency order, frequency order, and random order) for tasks involving acquisition of targets in a Zipf-like ¹ distribution and found that the stable layout was significantly faster than the recency layout, with performance benefits increasing with expertise. The authors evaluated the size morphing operation and found that gradual size ’morphing’

of item areas in the display did not affect performance - importantly, morphing did not substantially harm the participant’s spatial memory for target locations.

1http://en.wikipedia.org/wiki/Zipf’s law

Figure 2.11: SCOTZ

Figure 2.12: Size morphing 2.2.2.6 FST

FST technique has been proposed to view and manipulate hidden content through unimportant regions of overlapping windows Ishak & Feiner (2004), and using mouse-

2.3 Group Switching Techniques

over pie menu to display thumbnails of all windows that lie beneath that selected pixel.

However, the techniques only considered the windows close to the mouse cursor, and when the pie menu displays multi items, it becomes difficult to display thumbnails of each window at a recognizable size.