X hits on this document





6 / 10

phosphor versions of standard GUI widgets. Widgets are implemented as Flash components [12], which allows reus- ing them in other Flash applications.

Figure 13: Our Delphi-based prototype showing several phosphor transitions.

To objectively evaluate user performance using phosphor transitions we carried out two user studies.

USER STUDY 1: FOLLOWING A COLLABORATOR The purpose of the first study was to examine whether phosphor helps users visually track transitions in the user interface. The participant’s task was to watch a (simulated) collaborator adjust widgets in a control panel dialog. Then participants had to demonstrate their comprehension by undoing as many of the observed actions as possible. Our main hypothesis was that the phosphor interface condition would lead to better perception and memorization of the activities, which we would measure as faster task perform- ance.

Task Each trial took place in the following six steps. (1) Users clicked a “go” button. (2) A dialog appeared and the user’s mouse was disabled. The dialog contained a mix of sliders, combo boxes, and checkboxes laid out in a regular 5x8 grid as shown in Figure 14. Layout and labeling of widgets changed each trial. To indicate the non-interactivity, the window bar of the dialog was shown in gray. (3) After a pause of 500 milliseconds, the mouse pointer started trav- ersing the screen and consecutively adjusted six of the forty widgets in the grid, i.e., two sliders, two checkboxes, and two combo boxes. Simulating a human motion pattern, the pointer would hover 200ms over a widget before adjusting it. There were three playback speed conditions. In playback speed conditions “medium” the mouse traversal from one widget to the next took 600ms, 800ms, or 1000ms in ran- dom order, always resulting in the same overall duration of 4800 ms. In the “slow” and “fast” playback speed condi- tions all traversal and hover times were multiplied by fac- tors 1.3 and 0.7 respectively. (4) There was a 4000ms pause. At the end of that pause, all of the 4000ms afterglow effects had faded completely. (5) The mouse was re- enabled, the window bar of the dialog turned blue, and a text was displayed instructing participants to start undoing as many of the observed changes as possible. (6) The par-

ticipant adjusted as many widgets as they desired. Then they clicked the dialog’s “OK” button, which closed the dialog and completed the trial. Task time was counted from the moment the dialog turned interactive until the moment the user hit the “OK” button.

Interfaces There were two interface conditions, phosphor and control. The phosphor condition was identical to the control condi- tion, except that widgets displayed an afterglow after being adjusted by script or participant as shown in Figure 1. Each afterglow faded independently after 4000ms.

Figure 14: The apparatus used in the first user study. The user’s task was to undo as many of the six adjustments observed.

Experimental design The study design was within subjects 2 × 3 (user interface x playback speed) with 8 repetitions for each cell. For each trial, we recorded completion time and error. Error was any difference between the initial state before the automated traversal and the final state after the user adjustments; click- ing “OK” without making any adjustments, for example, resulting in six errors. Interface order and speed factors were counterbalanced. Dialog layout was randomized.

Participants received training upfront and at the beginning of each block. They filled in a questionnaire at the end of the study related to subjective preference, learnability of phosphor transitions, and usefulness. The study took about 20 min per participant.

Apparatus The experiment was run on a PC running WindowsXP with a 17” LCD monitor, at a resolution of 1280x1024 pixels. The interface used in this study was implemented in Mac- romedia Flash as described earlier. The optical Dell mouse was set to a medium mouse speed.

Participants 12 university students (1 female) between the ages of 24 and 30 participated in this study. All had experience with graphical user interfaces and mice and had normal or cor- rected to normal vision and color vision.

Document info
Document views33
Page views33
Page last viewedMon Jan 16 11:39:57 UTC 2017