The Effect of Zooming Speed on a Zoomable User Interface

Discussion

The goal of the experiment is to compare and analyze the zooming speed effect in a zoomable user interface in quantitative terms of performance time (seconds) for a set of searching tasks, as well as the recognition and retention accuracy. We also analyze the subjective preference.

Performance time

Our first hypothesis of the performance efficiency was proven to be true based on the statistical analysis of the performance time for the six zooming speeds. The chart of the performance time has a saddle shape with the longest time at the slowest zooming speed, 1.5 factor/second. The performance time is also long for the fastest zooming speed. There exists an optimum speed range as we expected. The shortest mean time appears at the zooming speed 8 factor/second in the middle of the zooming speed scope. The performance time at speed 4/sec and 12/sec also seems reasonably good. A one-way ANOVA was performed on the six treatments of zooming speed. The result (see in Appendix) showed that there exists a statistically significant difference among the means for different treatments at alpha = 0.01 (N = 204 and K = 6).

The slow zooming speed has worst efficiency without surprise. The reason is that subjects had to wait for a long time to zoomed in/out an object to allocate it. The time was longer if subjects went to a wrong part of the map. When the zooming speed was slow, the delay was long. This delay time forced subjects to wait when they found they did something wrong and wanted to zoom out and search again. The subjects survey also shows that subjects were dissatisfied with the slow zooming speed.

There are two possible reasons why performance time gets longer when the zooming speed is very fast (for our case, 12 factor/second and 16 factor/second). The first reason, which was also mentioned by the subjects, is that zooming effect is hard to be controlled using a mouse. Sometimes the subjects got lost because they zoomed too fast. The second reason is that too fast zooming speed wastes the machine resources. The memory and monitor cache can't keep track of the fast changes. The screen flicked greatly sometimes when the zooming speed was too fast.

There are many reasons for the deviation. Some subjects preferred panning instead of zooming, although it's rare, it could affect our data. Sometimes the subjects misunderstood the task and spent time in zooming in and out to make the correction, which made the performance time longer. The subjects' ability of searching in a graphic interface is an important factor for the performance, as shown in our raw data.

The chart of "mean browsing time for question sets" has statistically significant period with increasing browsing time for every 4 questions for a specific zooming speed. The reason of the period is that we organized every four questions to be a set of tasks, which were taken at the same zooming speed. The four questions were showed to subjects from easy ones to the difficult ones. The average difficulty of a task group was similar to the average difficulty of any other group.

To eliminate the bias effect, we let half of the subjects do the experiment with zooming speed going up and another half do the experiment with zooming speed going down. The subjects were assigned to use increasing speed or decreasing speed randomly. We find the browsing time is relatively longer for the first several questions, which is shown by our raw data. More training questions might reduce this bias effect.

Retention Accuracy

On the chart "Retention Accuracy", we record the location recovery rates for different zooming speeds. After the searching tasks, we asked each subject six questions related to the tasks they performed. For each question, score 1 is assigned for a correct answer, score 0.7 is assigned for an answer of close location, and zero for an answer that not even close. The goal is to test how well subjects can recall the locations on the map for different zooming speeds. Since each group of tasks (four questions each group) is mainly within one location, we can assume that the retention accuracy only depend on the zooming speed. However, the main goal of our design was to test subject's performance on a zoomable interface environment. We allow the subjects to have top-level view of the whole map, so subject may establish the long term memory of the locations in the previous tasks while doing the later tasks.

Over all, we didn't find any significant effect of the zooming speed on retention accuracy based on the retention question score chart and ANOVA statistics of retention questions. We observed the highest retention accuracy for slowest zooming rate, which corresponds to the first 4 questions if subjects used increasing zooming speed or the last 4 questions if subjects used decreasing zooming speed. There is no explainable pattern on the chart. The questions set for 8/second zooming speed seems hard to be recovered by subjects. We observed an obvious difference between the first 4 questions and the last 4 questions when they were given in the zooming speed level 16/sec. A better design strategy might divide subjects into groups for different zooming speeds.

Subjective Preference Survey

Most of our subjects preferred the high zooming speed rather than the slow speed. Some of them felt they are not sure what zooming speed they prefer or the zooming speed had little effect on them. The reason for this result was mentioned by our subjects. Almost all the subjects could not tolerate very slow zooming speeds because they had to wait for a long time to see the results of their actions. If they made mistakes, they could not recover quickly under the slow speeds. Even though fast zooming speeds was hard for subjects to control the zooming effect at first, many of them were able to control the zooming effect after answering one or two questions using the fast zooming speeds.

Department of Computer Science: Direct questions and comments to the student editorial team