Visualizing Directions and Schedules on Handheld Devices

Visualizing Directions and Schedules on Handheld Devices A Pilot Study of Maps vs. Text and Color vs. Monochrome

Pankaj Thakkar, Irina Ceaparu, Cemal Yilmaz {thakkar, irina, cyilmaz}@cs.umd.edu Department of Computer Science University of Maryland College Park December 10, 2001

Abstract hand-held information devices like Palm Pilot, iPAQ and Jornada [4]. Rover [1] is a location aware information service framework for handheld devices Rover [1] is a project at the University of being developed at the University of Maryland, College Park that is trying to put Maryland. It can be deployed in places like these technologies together to create a shopping malls, museums and schools. This framework for providing information. The project is aimed at developing the user information is customized according to the interface for an in-building Rover system to device profile, the user profile and the be deployed in the AV Williams building of location of the user. Similar projects are the Department of Computer Science. The being worked on in other academic labs and design is determined by different treatments industries. HP has designed CoolTown [5] of the system functionality. Some of the that integrates all the objects with the web functions that are part of the Rover are: by using web-signs. Cornell University navigation aid, scheduling, locating people, designed an E-guide to provide navigation and email/chat. Our study conducted with 20 aid to visitors. In order to support typical subjects focused on 2 different tasks applications the Rover Technology uses corresponding to navigation directions and several basic capabilities: un-tethered scheduling. We conducted tasks to compare wireless communication, seamless the use of colors vs. monochrome displays integration across different technologies, and map-based vs. text-based directions. audio and video streaming, personalization, The results show subjective preference for multilingual capabilities and high degree of maps and colors over text and monochrome security. The user interacts with the system respectively but the performance differences using a hand-held device such as a Palm are not statistically significant. Pilot or a Compaq iPAQ. Applications of Rover system include: automated guide for Introduction museums, shopping assistant for shopping malls, personal assistant for students and Background visitors on campus.

With the prevalence of hand held devices For the user interface various devices can be and wireless networks people are not used like Palm Pilots, Compaq iPAQs, required to sit in front of their desktops to Fujitsu Pen Computers. All these devices access information. Various wireless have different functionality and interface. network technologies like IETF 802.11 [2] Some of the important characteristics of and Blue-tooth [3] are making it possible to these devices that affect the design of access the information independent of the information providers like the Rover Client location. Many companies like Compaq, are monochromatic and color screens, ability Palm, and HP are developing customized to display text and graphics, sound capability, input method, etc. Earlier compatibility between the stimulus, the versions of Palm Pilots can only display cognitive processing, and the response monochromatic text. Some versions of required for the task. The second factor is iPAQ like H3600 series can display 12 bit the degree to which several pieces of colors, have a faster processor to handle the information can be presented in a graphics and can play 16-bit sound. Our simultaneous format. project is to design parts of the user interface of the Rover Client for some of these Tasks that require spatial processing in devices. Specifically we are looking into working memory can be best served by user interface for navigation aids and visual spatial displays and more poorly scheduling for a Compaq iPAQ H3600 served by textual displays [10]. Text-only series device. This paper describes the descriptions are notoriously inadequate for design of the user interface mentioned expressing complex spatial relationships above. [11].

Previous Research Comparing text and graphics in navigation display design was recently studied [7]. In Map versus Text Directions this study, participants who are pilots were asked to choose an airport farther from an Two different types of resources play an approaching storm front by using a touch important role in designing navigation sensitive display panel. The independent systems for handheld devices: technical variables were 1) nearest airport information resources of the machine and cognitive display mode (text-only display, enhanced- resources available to the user [6]. text display; displaying the orientation of the Technical resources cover all kinds of airport, map display), 2) map display mode limitations of the handheld devices such as (north up or track-up), 3) aircraft heading screen size, resolution, colors, and (north or south), and 4) turbulence (on or characteristic of communication medium. off). Dependent variables were decision The design should obey the technical time, decision accuracy, and navigational resources but also take into account the accuracy. The study showed the superiority users’ limited cognitive resources to decode of graphical over textual information display and understand a presentation. of nearest airport information. Pilots were significantly faster and accurate using the Presentation of route directions range from a map display than using either the text-only 3D walkthrough guided by a virtual scout display or enhanced-text display. accompanied by spoken text to simple text explanations indicating the directions [12]. Color versus No Color Displays This paper examines the effectiveness of map-based versus text-based directions in The use of color in HCI literature and terms of both speed and accuracy in finding related disciplines varies in its focus. Colors the targeted locations. have been used for improving performance in search-and-locate tasks; recall tasks, The effects of graphics on human decision judgment tasks, and retention tasks. performance are highly dependent upon the tasks [8]. Therefore the question is: What Benbasat, Dexter, and Todd [14] are the characteristics of the tasks that we demonstrated the effectiveness of color on can make use of to graphically present learning and information acquisition in information? Two factors might influence decision support systems. In this study, the effectiveness of graphically versus color was found helpful in situations where textually presented information [9]. The first the decision maker was under a time factor is the S-C-R compatibility: the constraint, and at early stage of problem solving heuristic. Variables

Hoadley [13] studied the effects of color on Independent Variables a decision maker’s ability to extract information from different graphical and We chose one independent variable for each tabular presentations. In this study, of the hypotheses. For the schedule participants were presented with business experiment we chose number of colors as time series data in four different formats: our independent variable with two table, line graph, bar graph, and pie chart. treatments. They were asked to extract information from Monochromatic Display: 2 colors (black, these presentations. Time of information white) extraction and accuracy of the extraction Color Display: 5 colors (black, red, green, were measured. The results indicated that yellow, orange) color improves time performance for tables, pie charts and bar graphs, and accuracy performance for pie charts and line graphs. These finding are consistent with the literature regarding the use of color with alphanumeric data [15].

Our experiment is a pilot study and is focused on two specific tasks on handheld device (a Compaq iPAQ). Previous studies analyzed a larger variety of treatments for the tasks we are investigating. From those we chose minimal yet relevant treatments: For the navigation aid experiment we chose 2D Maps vs. Text (12pt font) based display of directions as our independent directions and colored vs. monochromatic variable with two treatments schedule displays. Map Directions: Use map of the building with path marked over it to show the Experiment direction that the user has to follow. Text Directions: Use text-based directions Introduction and Hypothesis for guiding the user.

Our experiment is designed to determine the effect of colors and graphics in displaying information on handheld devices. Specifically we investigated the effect of colors in displaying the weekly schedules of professors and use of graphics in giving directions to people within the building. We hypothesize that: 1) Use of colors results in statistically significant differences in performance time and number of errors over monochromatic displays. 2) Map-based directions produce statistically significant differences in performance time and number of errors over text-based directions. We designed 4 information displays with all of the subjects were computer science possible combinations of independent majors. We conducted a within subjects variables as shown in the tables below. A experiment. No training in using the iPAQ sample screen for each treatment is shown was required for the subjects as their above. interaction with the device was minimal. Prior to the experiment, our subjects were Display Screen Display Type given instructions about the purpose of the Display1: (Color) Screen DirectionColor Type experiment, their tasks, and the procedure of 1: (Maps) MapDisplay based the experiment. All the subjects were 2: (No-color) directionsMonochromatic familiar with the building in general, but not 2: (Text) TextDisplay based with the floor on which we conducted the directions navigation task in particular. All the subjects Dependent Variables signed the consent forms.

We measured two dependent variables for Materials each of the subject: 1) Performance time: Time needed to The following materials were used to perform the task correctly. If the user conduct the experiment: made any error the task was not 1) The simulation of 4 display screens with interrupted and the time taken to rectify their corresponding update screens (20 the error was included in performance screens in total). A stopwatch with 0.1 time. second of accuracy was used for timing the 2) Number of errors: Errors made by the experiments. user during the task was recorded as 2) Pre and post questionnaires well. 3) Written and oral instructions The display screens were presented to the All the materials mentioned above are user in a random fashion. Subjects were attached in the appendices. timed and observed during the experiment. Procedure Pilot study results The subjects were given instructions and We conducted pilot studies with 4 subjects. consent forms at the beginning of the We improved the display of map directions experiments. All queries were answered based on the feedback received from our before the experiments. pilot subjects. The orientation of the map Before starting the experiment, subjects was changed from portrait to landscape were given a training task for the scheduling mode to make the maximum use of available part of the experiment. The task was to find screen space and thus show a higher the next available slot to meet with a resolution map. We colored the corridors professor, given the schedule of that yellow to distinguish them from the rest of professor for a week and a starting time the building. (e.g.“suppose it is Monday 11 am; when can you schedule a meeting?”). The task had 2 Subjects treatments: color and no color screens. Then, the subjects were presented with the We choose subjects who are comfortable screens for the schedule experiment. The with using computers. Twenty graduate subjects were required to perform a first task students from the University of Maryland similar to the training task, with two participated in our study. Our subjects different treatments (color and no color). consisted of 4 females and 16 males Subjects were shown the weekly schedule of between the age of 20 and 30. The majority a fictitious Professor X. They were given a starting time and ending time in the week iPAQ does not have enough memory to run and were supposed to report the number of a complete web browser, we used remote slots in which a meeting with the professor display capabilities of the X server to run the could be scheduled (e.g.“what is the total application from a laptop and export the number of slots available for a meeting display to the iPAQ. The machines (iPAQ between Tuesday 12 pm and Wednesday 5 and laptop) communicated using 802.11 ad- pm?”). The subjects could schedule a hoc mode. The display screen on the iPAQ meeting either in office hours or available was controlled by an application written in hours. Tcl-Tk that controlled the opera browser. A screen shot of the control application is For the second task the subjects were taken given below. to the second floor of the A.V. Williams Building in the University of Maryland, College Park. The subjects were supposed to reach a particular spot on that floor from a certain starting point. The subjects were told about the orientation of the building and their relative position (e.g.: you are facing south). Initial directions were provided at the starting point. The directions were discreetly updated when the user reached a checkpoint. We divided the path into five checkpoints with three additional checkpoints in case the user went on the wrong path. For the map directions, the path the user had to follow was shown with a blue line with a red dot indicating the destination. Since we had no location determination technology, the user’s location was assumed to be on the last segment of the path (starting from the Results destination). The subjects were told about this setup. As the user walked along the The experiment collected data about the path, the directions were updated to display tasks users were presented with and their the directions from the current checkpoint. answers to the post-questionnaire. The In case of map directions, this meant the results were gathered in MS Excel tables for path started from the current checkpoint to statistical analysis. the destination (red dot). As the user came closer to the destination, the path became shorter. For the text directions, the user was advised to read the first 2-3 directions only Tasks results as we kept updating the directions as the user reached a checkpoint. T-tests were performed separately on the 2 tasks: visualizing schedules with colors and A Compaq iPAQ was configured to run with no colors and following directions with LINUX (kernel version 2.4) on it. The maps and with text. For both tasks T-tests screens were represented as full screen mode were done for the 2 dependent variables: of the OPERA web browser. All the screens performance time and number of errors. The were written in HTML. Maps were shown in results for the scheduling task show that the GIF format. Arial 12 pt. text was chosen for performance time between working with displaying all the information. Since the color schedules vs. monochrome schedules Task3: M ean Tim e do not have a significant difference (t(19)=-0.26, p=0.79 two tail). 200

128.68 150 Task2: Mean Tim e 117.58 100

12 50 10 7.15 0 8 6.45

6 Maps Text

2 Task3: Mean Errors 0

NoColor Color 0.4 0.37 0.37

0.3 Task2: Mean Errors 0.2

0.35 0.30 0.1 0.3

0.25 0 0.20 0.2 Maps Text 0.15 0.1 Questionnaire results 0.05

0 Out of 20 subjects, 13 prefer map directions NoColor Color to text directions and 7 prefer text directions Also, the analysis done on the number of to map directions. The t-tests show the errors in the two cases proves no more following results: t(19)=0.59 and p=0.56 significant differences than those for the two tail. Out of 20 subjects, 16 prefer performance time (t(19)=-1.24, p=0.28 two colored schedules to monochrome ones, and tail). 4 prefer monochrome schedules to colored ones. The subjects were asked to rate on a The results for the directions task again do scale from 1 (difficult) to 9 (easy) the use of not prove significant differences between text vs. maps. We calculated the average of the performance times, although it seems to their answers, and the overall result shows a be some difference in favor of the use of text slight subjective (but not statistically (t(19)=0.12, p=0.90 two tail). significant) preference for maps: Average Average Also, unlike the scheduling task, the number Preference Maps Preference Text of errors for using maps and text turned out to be the same (t(19)=0, p=1 two tail). 6.95 6.55 The subjects rated on a scale from 1 (useless) to 9 (helpful) the use of colors for visualizing schedules. Again, the average 6.65 shows a slight subjective (but not statistically significant) preference for colors. Subjects were also asked to rate on a scale from 1 (adequate) to 9 (inadequate) the amount and type of colors used. The average of their answers is 4. All data and statistical analysis are presented . Subjects were more comfortable with in the Appendices. colored schedules once they became accustomed to the task and technology. Discussion . The size of the map and of the schedule was an issue, and can be improved. The goal of our experiment was to establish the importance of using maps vs. text in Conclusion giving directions and of using colors vs. no colors in displaying schedules. To achieve Impact for practitioners this goal we conducted separate tests for Our experiment tried to establish the each of the 2 tasks: following directions and effectiveness of maps over text for working with schedules. Our hypotheses navigation directions and effectiveness of were that, for directions, maps are more colors over monochromatic schedule efficient than text, and, for schedules, colors displays. The results of our experiments are more efficient than monochrome. We showed a subjective preference for maps and had 20 subjects for our experiments and colors though we could not get statistically after doing the statistical analysis, we significant differences to support our concluded that we needed more subjects in hypothesis. order to get statistically significant differences. Unfortunately, the results Suggestions for future research neither support nor contradict our From the experience of our studies we hypotheses. The data collected for the 2 derive the following suggestions for future tasks (performance time and number of research. errors) show no statistically significant . Increase number of subjects to achieve differences. However, the post-questionnaire statistically significant results. results indicate subjective preference for the . There are various treatments possible use of maps and for the use of colors. within the treatments we studied. For We conducted 1 task for schedules with 2 example one can use 3D maps, treatments: colors and no colors. We landmarks and symbols to give conducted 1 task for directions with 2 directions. treatments: maps and text. . Contrasting colors need to be used to As for the questionnaire, subjects were distinguish between different categories asked to evaluate the tasks and also to of schedules. provide some comments about the whole experiment. References The conclusions we derive from their comments are: [1] MIND Lab. “Rover System.” Work . Text based directions were a little in progress. confusing in the way they were updated as the subject was moving towards the [2] IEEE 802.11 Working Group. “802.11 destination. Wireless Network Standard.” November . Subjects’ opinions about updating 2001; http://grouper.ieee.org/groups/802/11/ directions were split between updating too fast and updating too slow. [3] The Official Bluetooth Wireless Info Site . The colors used for the schedule would November 2001; http://www.bluetooth.com have been more helpful if they were chosen from a larger variety of colors [4] Compaq Research Labs. “Compaq (more contrast). iPAQ.” November 2001; http://www.compaq.com/ipaq presentation: An integration of the [5] HP Research Labs. “HP CoolTown.” findings.” Communications of ACM, 29(11), November 2001; http://cooltown.hp.com 1094-1105, 1986.

[6] Butz, A., Baus, J., Kruger, A., Marco [15] Christ, R.E. “Review and analysis of Lohse. “A hybrid indoor navigation color coding research for visual displays.” system.” Intelligent User Interfaces, 25-32, Human Factors, 17(6), 542-570, 1975. January 14-17, 2001

[7] Williams, W. Kevin. “Comparing text Acknowledgements and graphics in navigation display design.” Federal Aviation Administration, February We thank our professor Dr. Ben 2000 Shneiderman, Founding director of HCIL, University of Maryland at College Park, for [8] Tullis, T.S. “An evaluation of working closely with us in each step of our alphanumeric, graphic, and color project. His guidance helped us go in the information displays.” Human Factors, right direction. We also wish to thank Dr. 23(5), 541-50, 1981. Ashoka Agrawala from the Department of Computer Science, University of Maryland [9] Wickens, C.D., Scott, B.D.. “A at College Park, for his advice and technical comparison of verbal and graphical support. We thank all of our subjects for information presentation in a complex participating in the experiment during their information integration decision task.” Tech. busy schedule. Report EPL-83-1/ONR-83-1), University of Illinois at Urbana Champaign, June 1983. Credits [10] Wickens, C.D., Sandry, D.L., Vidulich, M. “Compatibility and resource competition Pankaj Thakkar between modalities of input, output, and Project proposal, experiment design, , pilot central processing: Testing in a model of study, recruiting subjects and running the complex task performance.” Human experiment, introduction writing, results and Factors¸ 25(2), 227-48, 1983. conclusion writing

[11] Towns, S., Callaway, C., Lester. Irina Ceaparu “Generating coordinated natural language Project proposal, pilot study, running the and 3D animations for complex spatial experiment, results, discussion and explanations.” AAAI-98: Proceedings of the conclusion writing 15th National Conference on Artificial Intelligence, 1998. Cemal Yilmaz Project proposal, experiment design, [12] Kruger, A., Baus, J., Butz, A. “Smart reference research, questionnaire design, graphics in adaptive descriptions.” AVI pilot study, recruiting subjects and running 2000, 92-97, 2000. the experiment, background writing, conclusion writing [13] Hoadley, D.E., “Investigating the effects of color.” Communications of ACM¸ 33(2), 120-139, 1990. Appendices [14] Benbasat, I., Dexter, S.A., Todd, P. “An 1. Pre Questionnaire experimental program investigating color- 2. Post Questionnaire enhanced and graphical information 3. Task Results 4. Questionnaire Results 5. Statistics