The Effect of Zooming Speed in a Zoomable User Interface

Experiment

Introduction and Hypothesis

This experiment sought to measure the effect of zooming speed to the user performance and the perception of users when users perform searching tasks. The overall goal was to find the best zooming speed at which user performance is highest and the retention of information is best.

Hypothesis 1: There will exist a zooming speed at which the performance time is shortest. Performance time will increase if the zooming speed is too low or too high.

When zooming speed decreases, the time used to zoom from one scene to another will increase. Users will have to wait longer time before they are able to get results from system and evaluate the results. Extremely fast zooming speed, on the other hand, will lead to very short response time, which may result in high user anxiety and error rate. In a zoomable User Interface, when the zooming speed is too high, users will feel difficult to catch some important intermediate states of the information searching process.

Hypothesis 2: The retention of information will improve as zooming speed decreases.

Hypothesis 3:Subject preference will decrease if the zooming speed is too low or too high. Most subjects will prefer the zooming speed at which performance time is shortest

Independent Variable: Zooming Speed (6 treatments)

1. 1.5 scale factor/ second
2. 2 scale factor/ second
3. 4 scale factor/ second
4. 8 scale factor/ second
5. 12 scale factor/ second
6. 16 scale factor/ second

Dependent Variables:

1. Performance time: time spent to complete search tasks
2. Retention Accuracy: accuracy rate to answer the retention questions
3. Subjective Preference

Pilot Study Results

Four subjects participated in the pilot study. It took each subject about five minutes to answer the subjective questionnaire, read the experiment introduction, and finish the training questions. Answering the twenty-four searching questions and the subjective satisfaction survey question took each subject eighteen to twenty-three minutes. Answering the six retention test questions took each subject about 2 minutes. The pilot study pointed out several flaws in the overall design.

Here is a list of changes that we had made based on the pilot study.

1. One more training question was added to ensure the participant be familiar with the user interface. Originally we had one training question and it turned out to be insufficient. Subjects still had some difficulty in using the experiment software when they performed the first task.
2. The experiment software was modified slightly to measure the time more precisely. The pilot study showed that the time subjects used to read and the questions took significant portion in the entire time used to answer the questions. We redesigned our software timer to start after subjects finish reading the questions.
3. Questions were specified more clearly. Many times in the pilot study, participants could not find the correct answer because the questions were misunderstood.
4. The sizes of the experiment windows were fixed so that subjects had no chance to change it.
5. We modified our software to save experiment data each time when the subject answered one question. Originally the data were saved only when the experiment windows were closed. We lost the experiment data of one subject because the system crashed after the subject finished all the questions.
6. Error rate was dropped from the dependent variables list. Error rate was originally one of the dependent variables. The pilot study showed that there was no significant difference of error rate. The experiment allows subjects to use as much time as they want to answer the questions. All the pilot study participants were able to correctly answer each question.

Subjects

Thirty-nine subjects participated the experiment and five were dropped because the experiment was not finished due to various reasons such as the program crashed due to low virtual memory. All subjects participated on a voluntary basis. Among the remaining Thirty-four subjects, ten were females and twenty-four were males. The ages of the subjects ranged from younger than 18 to older than 40 years. Nineteen were in the age range from 18 to 25. Nine were in the age range from 26 to 40. Two were older than 40 and four were younger than 18. Twenty-five were students at the University of Maryland and nine were professionals. The majority of the participants (twenty-two) were majoring in Computer Science, three were majoring in Electronic and Computer Engineering and the remaining were in a variety of disciplines (e.g. Mathematics, Physics, Business). Most of the participants (twenty-nine) used computer regularly (more than 2 hours per day). Only one subject had previous experience with JAZZ. None of the subjects had ever been to Tokyo Disney land that has some similar layout as our information space.

Materials

Software: The experiment software is based on a zoomable user interface JAZZ which was developed at HCI lab in University of Maryland at College Park. The detailed information about JAZZ can be found at http://www.cs.umd.edu/hcil/jazz/. The experiment software was implemented by modifying the code in JAZZ that is related to zooming speed and by extending the demonstration application (HiNote) of JAZZ.

JAZZ does not support different zooming speeds. All the zoom operations in JAZZ are smooth zooming. We modified some JAZZ functions to support different zooming speeds. Our information space was a tour guide map of an imaginary Disney land--"Fantasy Land" showed in Figure 1 below. The map was designed and displayed in HiNote. A question window (Figure 2 and 3) is displayed on the right side of the original HiNote window. The question window was used to show instructions, training questions, the twenty-four search questions and one subjective satisfaction survey question. All the code was written using Java 1.2.2.. Figure 1 is a snapshot left-hand side window of the experiment with a top level view of the fantasy land when the experiment starts. Figure 2 showed the right window that is used to be display the instructions and explanations.

Figure 1. Top level view of the fantasy land

Figure 2. Start up window with instructions

The information provided in the tour guide map of Fantasy Land (in the right window, Figure 1) has different levels. Subjects have to zoom in or zoom out to see the information contained in another level. Twenty-four questions (in appendix A3) were designed to measure the user performance. Each question was shown as in Figure 3 - the warm-up question window and it requires users to search certain information in the Fantasy Land by zooming in or zooming out. For each treatment, there are four search questions. Six questions were designed to measure the retention rate. There is one subjective satisfaction survey question. The software automatically records the answer to each question except the six retention accuracy questions. The data were saved automatically into a file. The six retention accuracy questions were answered using paper-based form.

Figure 3. Warm-up question

Experimental Setting and Hardware: All sessions of experiments were conducted in the Sun Ultra 5 computer in Junkfood lab. The experiment software was installed in three Altra 5 computer and three subjects could run the experiment at the same time. All the computers had the same type of monitor.

Paper-Based Forms: Three paper-based forms were prepared for the experiment: consent forms, background survey, and the forms for the six retention questions.

The six retention questions were used to measure the retention accuracy. Assigning 1 point for a correct answer, 0.7 point for the answer that is nearly correct and 0 point for a totally wrong answer.

Procedures and Problems

A typical session of experiment has the following five steps:

1. The subject reads and signs consent forms.
2. The subject fills out background survey (as in appendix A1.
3. The subject reads the instruction (as in appendex A2 and Figure 2) and practices using the mouse to zoom in and zoom out.
4. The subject performs the training tasks and asks all the clarification questions he/she might have.
5. The subject answers the twenty-four searching questions (as in appendix A3) by zooming in or out in the information space.
6. The subject answers the satisfaction survey question (as in appendix A4).
7. The subject answers the six retention questions(as in appendix A5).

When the subject started the experiment, there were always two windows on the screen. One window showed the fantasy land as in Figure 1, 4 and 5 on the left hand side. Another window showed the searching task questions and instructions on the right hand side. The subject started with windows shown as Figure 1 and Figure 2. After the subjects carefully read the instruction, they would click the 'START' button and the first warm-up question would show to the subjects as Figure 3. First, each subject needed to read the question carefully and then clicked 'let me go' button. The window with the fantasy land was frozen until the subject clicked the 'let me go' button. Then the subject was enable to manipulate the map and question choices. Figure 4 and 5 showed the screen shots during the task. Time that each subject spent to answer each of the twenty-four search questions were recorded. The questions were showed in the same order for each subject. Half of the subjects conducted the experiment from low speed to high speed. Another half of the subjects conducted the experiment from high speed to low speed.

Figure 4. The attraction 3 for the first warm-up question


Figure 5. The description with ticket type, etc.

The six retention questions were used to measure the retention accuracy. Each question corresponded to one zooming speed. 1 point was assigned for a correct answer, 0.7 point for the answer that is nearly correct and 0 point for a totally wrong answer.

The main problem during the experiment was that the program requires much memory if the tour guide map is zoomed in too much. A few times, the program crashed after a subject answered more than half of the questions. Five participants were thus dropped from the analysis. Another problem was that a couple of subjects tended to rely heavily on panning to search the information space. They were not comfortable with the action of pressing right button of the mouse and dragging, which is the zooming mechanism used by HiNote. We took this into consideration when we analyzed the data.

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