Home

0 - Deep Blue - University of Michigan

1. 331 35 335 355 CN CO wrr ve e ve vre ere CD TWO N 00 74 MAIN EXPERIMENT We will now move on to the next part of the study e Walk the subject to the Sim Lab e Turn on the power strip of the sound cart and push 2 3 eff and autotake e Insert the tape e Push record e Make sure the counter is advancing and the picture is clear e Turn on the overhead and PC projector gt Set up the car for the next block according to the pattern list where Light sunroof and dashboard lights and Dark headliner light e Have the subject get in the car and if necessary show him how to adjust the seat Just before using the simulated Ali Scout e Run the Touchscreen control panel and have the subject calibrate the touchscreen e Run Touchscreen Trainer to make sure the calibration was correct e Put the frame on the touchscreen e Open the simulation program and fill in the appropriate information output file DE S R CX e Turn off the room lights When hand you a card please use the system to find the location listed on the card When you have found it please press the FOUND key then the cancel key the circle and put the card in the envelope will then give you another card Do you have any questions Please turn on the system e Run the retrieval series The next phase involves entering some locations into th
2. For simulated interfaces it is important that a very fast computer be provided In fact earlier versions of the simulation ran on Mac llcx and were noticeably slow While running the simulation on a PowerMac 7200 and optimizing the code somewhat improved performance use of a still faster computer would have been closer to reality since subjects were still noticeably slowed down when trying to repeat the same keystroke several times in a row Such computers are now widely available Developers planning to use rapid prototypes for interface evaluation need to make sure that computers are available that are fast enough to run simulations without perceptible delays especially in response to keystrokes As computer performance improves this is becoming less of an issue However the tendency is to continue to challenge the boundaries of computer performance by writing more complex simulations In addition to improving system timing use of the real interface could be enhanced by providing keys with more distinctive tactile feedback to indicate closure In the case of simulated interfaces some thought needs to be given as to how readily fabricated collapsible overlays might be used to provide feedback An alternative might be to attach a small vibrator to the subject s fingertip connected to the simulation computer that would give the feeling of switch closure Problem 11 The hinged panel was not stable When entering data subjects used a
3. P Value Subject Sex Age Grour 107579 594 3361 862 4 984 0001 12267 156 12267 156 18 186 0001 Condition Age Group 2 15874079 7937039 11 767 0001 320 215852002 6745388 ANOVA summary of condition effect and some interactions for destination retrieval The dependent variable is retrieval time Source df Sum of Squares Mean Square F Value P Value 496 75377700 151971 2 ANOVA summary of condition effect and some interactions for destination entry The dependent variable is entry time Mean Square F Value P Value Source df Sum of Squares Condition 2 anraaze 18589438 19495 0001 0001 1990 Condition Age Group al 20026051 5006 513 5251 0004 95359 472950 268 88
4. How did performance with the Ali Scout interface compare with other systems described in the 54 I CIOSIIAG ME u 56 Vii REFERENCES APPENDIX A APPENDIX B APPENDIX C APPENDIX D APPENDIX E APPENDIX F APPENDIX G APPENDIX H APPENDIX 1 APPENDIX 4 APPENDIX K 57 MEAN TASK TIMES FOR MOTOROLA PROTOTYPE Dira 61 DISPLAY UNIT 63 LIGHT LEVELS FOR THE EXPERIMENTAL CONDITIONS usce opa TRE 65 BIOGRAPHICAL 67 CONSENT FORM eene 69 INSTRUCTIONS TO 0 71 EXPERIMENTAL 11 79 PRACTICE SHEET einn anms eie 81 RETRIEVAL 83 DISPLAY UNIT POST TEST QUESTIONNAIRE 85 ANOVA TABLES ae 87 viii INTRODUCTION Why this topic is of interest One of the more interesting recent developments for motor vehicles has been the advent of navigation systems Most of the systems planned or in production identify the location of a vehicle on an electronic map and provide both visual and voice directions to driver selected destinations In some systems current traffic conditions are considered in route calculations Navigation systems are very popular in Japan Treece 1996 and may see broad market penetration
5. This document is driving navigation route guidance available to the public through the cartography usability input devices National Technical Information Service controls keyboards Sprinafield Virginia 22161 19 Security Classify of this report 20 Secunty Classify of this page 21 No of pages 22 Price none none 91 Form DOT F 1700 7 8 72 Reproduction of completed page authorized Destination Entry and Retrieval with the Ali Scout Navigation System HUMAN FACTORS FACTORS UMTRI Technical Report 96 30 Al B2 03 04 ES 67 19 JO K Aaron Steinfeld Daniel Manes See oo Paul Green David Hunter Lu H N 0 S T UU HX YZ University of Michigan 2 A O F amp Ann Arbor Michigan USA The Ali Scout Keyboard 95 actual size ae ISSUES 1 How long does it take to enter and retrieve destinations using the Ali Scout 2 How does the Ali Scout compare with other navigation systems 3 How does performance vary as a function of driver age and sex ambient illumination dusk vs night and interface type real vs simulated 4 What kinds of problems do drivers encounter and how can they be corrected 5 How accurate are subjects in looking up coordinates in the manual Simulated Real Ali Scout Ali Scout Retrieve destination from units PE M HN E nter new destination into memor ADA ROE
6. the navigation system provided a means for directly recording keystroke entries and Screen changes Navigation hardware developers should incorporate such a capability in development systems The absence of such features discourages the conduct of human factors studies leading to systems that are more difficult and less safe to use What were typical entry and retrieval times and error rates for destinations The mean entry time was 65 seconds the median was 52 seconds The mean retrieval time was 10 seconds the median was 6 seconds Thus entry times were 6 to 9 times greater than retrieval times for stored destinations The difference between the means and medians was so great because of the presence of several extremely long times which influence the mean more than the median Because of these extreme outliers the median gives a better sense of a typical time than the mean It should be recalled however that these times do not include the subject s thinking time required for planning and confirmation By observation those times were typically 1 to 2 seconds for entry and 1 to 2 seconds for retrieval If the subject was confused these times could be much longer Further for destination entry a significant time was required to find coordinates on lists or on maps a point discussed later in this section Error rates were 1 5 percent for retrieval 10 percent for entry nearly a factor of 7 difference Did performance change wit
7. 5 percent but 54 errors for entry 10 0 percent An uncorrected error is an instance in which the subject continued with the experiment believing that the information on the card had been correctly retrieved or entered when in fact the 32 destination had not been correctly retrieved or entered Thus an error was counted only if the end result was incorrect not if the subject typed incorrect keys and then corrected them while completing the task More will be said about errors in the next section What were typical entry and retrieval times and error rates for destinations Table 15 summarizes the median maximum and for the sake of completeness mean for each of the cards and Figure 20 gives a graphical depiction of the median and mean values for each destination to give a sense of their variability Table 15 Median Maximum and Mean times for each of the 30 cards Hetrieval Times s Place Name SAKURA BANK BIR ICE ARENA MONTERREY REST MOBIL BIG BOY SEAFOOD BAY PRINT GALLERY MAJESTIC CAFE VANDENBURG SCH BIR LIBRARY BILL KNAPPS PRIMOS PIZZA WOODSIDE HOSP BIR THEATER MONGOLIAN BBQ Overall 3 47 12 00 9 58 3 98 4 57 8 27 6 82 2 37 3 56 9 97 4 79 6 15 0 38 9 83 7 27 6 23 101 88 121 75 120 58 55 40 51 83 94 90 16 05 21 30 52 97 58 13 62 60 57 63 19 35 30 77 39 77 121 75 Median Maximum Mean 8 80 19 00 16 34 8 68 7 47 17 10 7 52 3 73 9 68 14 05 8 70 10 93 1 8
8. Evaluation Technical Report UMTRI 93 45 Ann Arbor The University of Michigan Transportation Research Institute Sears A Revis D Swatski J Crittenden R and Shneiderman B 1993 Investigating Touchscreen Typing the Effect of Keyboard Size on Typing Speed Behaviour and Information Technology 12 1 17 22 Siemens Automotive undated Ali Scout Navigation System User s Guide Auburn Hills MI Siemens Automotive Segal L D and Andre A D 1993 Activity Catalog Tool A C T v2 0 User Manual NASA Contractor Report CR 177634 Moffett Field CA NASA Ames Research Center Steinfeld A Manes D Green P and Hunter D 1996 Destination Entry and Retrieval with the Ali Scout Navigation System Technical Report UMTRI 96 30 Ann Arbor MI The University of Michigan Transportation Research Institute this report Taylor W and Wu J 1995 A Database System Containing MOE s of Interest to the Evaluation Technical Report EECS ITS LAB FT95 028 Ann Arbor MI The University of Michigan Intelligent Transportation Center Treece J B 1996 In Japan Car Buyers Put Navigation Devices at the Top of Their List Automotive News September 16 1996 p 3 24 Underwood S E 1994 FAST TRAC Evaluating an Integrated Intelligent Vehicle Highway System Proceedings of the IVHS America 1994 Annual Meeting volume 1 300 311 Washington D C IVHS America Wallace R R Eby D W and Gardner S
9. F Loring B A and Wiklund M E 1991a Test of Reduced Size Touch Screen Keyboards technical report Bedford MA American Institutes for Research Coleman M F Loring B A and Wiklund M E 1991b User Performance on Typing Tasks Involving Reduced Size Touch Screen Keyboards Vehicle Navigation and Information Systems Conference Proceedings VNIS 91 New York Institute of Electrical and Electronics Engineers 534 549 Detweiler M C 1990 Alphabetic Input on a Telephone Keypad Proceedings of the Human Factors Society 34th Annual Meeting Santa Monica CA Human Factors Society 212 216 Dingus T A Hulse M C Krage M K Szczublewski F E and Berry P 1991 A Usability Evaluation of Navigation and Information System Functions SAE paper 912794 VNIS 91 Proceedings Warrendale PA Society of Automotive Engineers 527 536 Eby D W Streff F M Wallace R R Kostyniuk L P Hopp M L and Underwood 5 1996 An Evaluation of User Perceptions and Behaviors of FAST TRAC Pilot Study Results Technical Report UMTRI 96 14 Ann Arbor MI The University of Michigan Transportation Research Institute Green P Boreczky J and Kim S 1990 Applications of Rapid Prototyping to Control and Display Design SAE paper 4900470 Special Publication SP 809 Warrendale PA Society of Automotive Engineers Green P Levison W Paelke G and Serafin C 1993 Preliminary Human
10. Factors Guidelines for Driver Information Systems Technical Report UMTRI 93 21 Ann Arbor MI The University of Michigan Transportation Research Institute also published as FHWA RD 94 087 McLean VA U S Department of Transportation Federal Highway Administration December 1995 57 Green P and Olson A 1996 Practical Aspects of Prototyping Instrument Clusters SAE paper 960532 Warrendale PA Society of Automotive Engineers Hoffmann ER Tsang K K and Mu A 1995 Data Entry Keyboard Geometry and Keying Movement Times Ergonomics 38 5 940 950 Katz S Fleming J Hunter D R Green P and Damouth D 1996 On the Road Human Factors Evaluation of the Ali Scout Navigation System Technical Report UMTRI 96 32 Ann Arbor MI The University of Michigan Transportation Research Institute in preparation Katz S Green P and Fleming J 1995 Calibration and Baseline Driving Data for the UMTHI Driver Interface Research Vehicle Technical Report UMTRI 95 2 Ann Arbor MI The University of Michigan Transportation Research Institute Kieras D 1988 Towards a Practical GOMS Model Methodology for User Interface Design chapter 7 in M Helander ed Handbook of Human Computer Interaction New York Elsevier Science Kostyniuk L and Eby S 1996 Natural Use Yoke Study Survey EECS ITS LAB FT95 018 Ann Arbor MI The University of Michigan Intelligent Transportation Systems Center Lor
11. G 1995 edited version of FAST TRAC VHS videotape 5 min Ann Arbor MI The University of Michigan Transportation Research Institute Zwahlen H T Adams C C Jr and DeBald D P 1988 Safety Aspects of CRT Touch Panel Controls in Automobiles in Gale A G Freeman M H Haslegrave C M Smith P and Taylor S P Vision in Vehicles Il Amsterdam Netherlands Elsevier Science 335 344 Zwahlen H T and DeBald D P 1986 Safety Aspects of Sophisticated In Vehicle Information Displays and Controls Proceedings of the Human Factors Society 30th Annual Meeting Santa Monica The Human Factors Society 256 260 99 APPENDIX Exploration Route Planning While Driving Other Tasks A MEAN TASK TIMES FOR MOTOROLA PROTOTYPE 2 Task Mean find out what the system can do 5 75 find out what the system thinks is the current address 20 14 find out your current compass heading 4 22 reset your present location to 0 Motorola Road specify 117 33 the full address find out you current location on a map 12 56 scroll the map down until you see Dundee Road 9 63 zoom out until you can see Interstate 294 labeled 19 50 change map view from N up to heading up and back 9 89 display restaurants on the map 40 56 find out where the nearest restaurant is 7 88 find out the address of the nearest restaurant 72 33 adjust the volume 32 75 set 542 Lindbert Lane as destination Pick it from a map 86 56 Tell the device t
12. Pe AMIA RAI Enter KROGERS with coordinates poen 422908N Scroll list Type aly DEE ra Subject ha has typing ANTE Subject has typed the destination name em KRUGERS 08322504 4229088 Subject has typed the coordinates NI Once the I is typed the rest of the name appears Destination Retrieval Median 6 23 5 Mean 10 48 s Destination Retrieval p Age lt 0001 16 p Sex 0055 p AxS 0388 Young Middle Older p Dsk Nt 3917 p Dsk Sim lt 0001 p Nt Sim lt 0001 Night Dusk Sim 5955 255 TEE dio ICT VON SENI SS Cn tn A DR REST CO o AP V RN en ccn s Hart MEDI ES n CONCLUSIONS Destination Entry 90 p Age lt 0001 p Sex 0001 p AxS 0948 Young Middle Older p Dsk Nt 0004 p Dsk Sim lt 0001 Nt Sim 0087 Time s Night Dusk Sim O on E RES 55252 SANNES Destination Entry e Median 51 48 s e Mean 64 68 s o a Women Destination Entry p Age 0001 p Sex 2799 p AxS 7867 Young Middle Older p Dsk Nt 5013 p Dsk Sim 0055 p Nt Sim 0316 Night Dusk Sim 2 Destination entry typically took subjects almost 60 5 excluding coordinate lookup time 30 to over 60 s while retrieval typically took under 10 s e Entry and retrieval times were significantly longer for older vs
13. SEARS To enter a destination the subject entered in the name first up to 14 characters including spaces Keying was somewhat confusing Notice that many of the keys have two characters on them The right character was shown in white the left in yellow To type the left character the subject first pressed the yellow up arrow key the lower left key on the keypad and then the key of interest So to type a period the sequence would be YellowUpArrow MPeriod To get to the numeric field they pressed the diamond key In the event that all 14 characters were used the cursor moved automatically The next task was to key in the longitude and latitude of the destination either obtained from a map or from lists of street address ranges and coordinates Additional details of the entry process are provided later Calibration and other functions were also available but they were not examined Figure 16 depicts the complete menu of user accessible functions The location entry tasks used the New Destination branch of the menu Figure 16 and terminated at the Input Coordinates node Since the experiment only used portions of the Ali Scout system the Actual Position submethod was not allowed The equipment to determine current position was disconnected Also subjects were not asked to do any of the Special Features tasks However they were not prevented from entering this part of the menu 21 Complete Menu 1 Power ON OFF
14. YZ oO 4 A V b E Y gt Figure 31 The Ali Scout navigation system Problem 1 The most frequently observed usability problem was that almost all subjects confused the zero button with the letter O button because the symbols representing them are virtually identical except for color This was not a size 44 problem The mean subjective rating of text size where 1 too small 2 just right 3 too large was 1 9 for the real unit 1 7 for the simulation even though the character sizes were identical Large age differences were not apparent in the data Problem 2 Subjects had trouble locating and understanding the function of the shift key Often subjects simply forgot to use the shift key when it was required Also in darker conditions many of the older subjects were unable to detect a difference between yellow and white characters the shift key was used for characters labeled in yellow leaving the location of the symbol on the key as the only cue Furthermore the yellow upward pointing arrow used to represent the shift key was ambiguous for some subjects as they tried to use the shift key to move the cursor up a line or were unable to find any shift key when learning to use the system Finally several subjects tried to hold the shift key down while typing the character as with a typewriter or pushed the shift key twice because they were unsure whether their first attempt worked neither of which produced th
15. all unwanted destinations from the Ali Scout and select AT THE START as the current destination e Cross off the current subject on the coordinate and entry retrieval patterns e Lock the door and return the key to the desk 343B e Staple all the paperwork and put it in the file 78 APPENDIX G EXPERIMENTAL DETAILS Coordinate Search Task Counterbalancing Subject Sex Age Format List Order Order 1 F yung IM A 1 2 3 2 F middle P 3 1 2 3 F old A P 2 3 1 5 F middle P M 2 3 1 6 F old A P Mil 2 3 8 F midde P M A 1 2 3 9 F old A P 1 2 M young 3 1 2 old A P 1 2 3 M young M A 2 3 1 M middle P M A 1 2 3 M old A P MIS 1 2 M young P 1 2 3 M middle P 1 2 M old A P 2 3 1 repeat for 19 36 Note Format order codes are P Points of Interest given the name find the coordinates in a list A Address Ranges given an address find the coordinates in a list M Ali Scout Maps given two intersecting roads find the coordinates on a map List order is a code for what subjects saw in the first second and third blocks Balancing Outcome for the Entry Tasks Average Average 1st Location 2nd Location 79 Counterbalanced Sequence for Conditions Demographics Block Order with Conditions Group 1 Sim A Night C Dusk B 1 Night B Dusk A Sim C 1 S
16. finger width where interkey width gt finger width or 2 center to center distance key width finger width To examine the effect of marking the keys the target key was indicated with a black spot An additional 10 young men served as subjects Only combinations of 1 3 or 5 keys were examined Key sizes and spacings remained the same For the full data set 95 percent of the variance in movement time was accounted for by the following expression MT 169 3 45 ID 2 10 Nkey In summary the key studies Paelke Coleman Detweiler etc suggest the following E If a touchscreen keyboard is to be used it should follow a Qwerty format though the exact spacing standard Qwerty Qwerty matrix does not matter For reduced resolution touchscreens the rank order of interface designs from best to worst is phonepad Qwerty scrolling list Zexel and double press TravTek The rank order of actual keying times and GOMS predicted keying times are the same for four representative interfaces However prediction errors of actual performance times can be considerable Except for omitting apostrophes there are no majority stereotypes for how drivers deal with unusual names those with q z or hyphens in them when using phone keypads for entry Of the methods for using phone keypads for name entry there is no single method whose performance is clearly superior to others though the repeat key method is best I
17. for destination entry were 90 to 120 seconds Destination retrieval appears to be faster than other interfaces while destination entry appears to be much slower Additional effort needs to be given to describing test conditions in sufficient detail data base content device method of operation response timing subject selection so that replication is possible This is often not the case in similar studies Differences in time due to gender were on the order of 20 percent with women being faster Age differences were approximately a factor of 2 with older subjects having problems with the interface under the night condition Older subjects should be the sample for legibility assessments Entry times for simulated interfaces were much longer than those for real interfaces though the general pattern of results was the same Methods for providing tactile feedback the feeling of switch closure need to be explored This may occur as a consequence of research on virtual reality Beyond this experiment one of the major topics of current discussion is what drivers should be allowed to do while driving It has been suggested that drivers should not be allowed to perform any destination entry or destination designation tasks while driving This experiment shows that there are large differences between retrieving a stored destination and entering a destination to store Furthermore the literature suggests that there are large differences due to retriev
18. for retrieval but comparable for entry Dingus Hulse Krage Szczublewski and Berry 1991 report that entering an unfamiliar destination required 130 seconds storing a route took 160 seconds and retrieving a stored destination took 50 seconds Since the Dingus et al data is for using a simulated interface written in SuperCard the comparable times for this experiment also for using a simulation are 15 seconds for retrieval and 74 seconds for entry It is not apparent whether thinking time was included in the tasks times Thus these times are slightly longer for retrieval but much less for destination entry Paelke and Green 1993 provide time and error data for entry of addresses In contrast to the work of Paelke and Green the destinations examined here were all in a single county so entering a city was not required Furthermore in contrast to real systems there was no need to enter a destination mode prior to beginning destination entry Overall times the mean of the driving and parked conditions of 43 seconds were reported for the phonepad 44 seconds for the Qwerty interface 55 seconds for the scrolling list similar to Zexel Rockwell Pathmaster Siemens Tetrastar and 76 seconds for the double press interface similar to the TravTek interface Times for parked conditions were 10 to 20 percent less than those for driving It should be noted that the driving task was of low fidelity and subjects did not devote the attention to t
19. groups See Table 12 Locations were ordered 25 so that the minimum number of keystrokes averaged across groups by trial was just over three Balancing in this manner facilitates looking at practice effects across subjects The minimum number of keystrokes for each location and the locations chosen are shown in Tables 12 and 13 Note that the sets are slightly unbalanced averages shown at the bottom of the table After the experiment was run an error was discovered in computing the minimum number of keypresses to reach the first four locations A down or up arrow must be pressed to enter the scroll mode The locations that were most readily found through the scroll mode were affected by this discovery In addition the location of VANDENBURG SCH Was erroneously believed to require only one keystroke to reach A shift activation was mistakenly ignored The locations in Table 12 that are shown in bold were originally believed to require one less keystroke Table 12 Minimum Number of Keystrokes for Retrieval Stimulus Set SAKURA BANK SEAFOOD BAY BILL KNAPPS MONTGMRY WARD BIR ICE ARENA PRINT GALLERY PRIMOS PIZZA ROYAL OAK DELI MONTERREY REST MAJESTIC CAFE WOODSIDE HOSP SEARS VANDENBURG SCH BIR THEATER BIR ART GALLRY BIR LIBRARY MONGOLIAN BBO PALACE OF In a manner similar to the retrieval tasks an effort was made to equalize the total number of keystrokes and shifts across orders and across entry trials to facilitate
20. in the U S Such systems can reduce wasted travel saving drivers time and fuel and provide for operational efficiency by optimizing use of the road network By decreasing driving under uncertain conditions and eliminating the use of paper maps while driving accidents may be reduced Finally these systems will offer comfort and convenience to drivers However such positive outcomes are predicated upon the assumption that these systems are safe and easy to use There are two primary driver tasks in using these products 1 entering and retrieving stored destinations and 2 following the directions given by these systems route guidance Secondary tasks include setting and calibrating the system Route following deserves the most attention because it occurs while the vehicle is in motion Route following is covered in other reports in this project Katz Green and Fleming 1995 Katz Fleming Hunter Green and Damouth 1996 However destination designation also must be considered Generally destination designation is assumed to be performed while the vehicle is stopped or parked However in many circumstances such as driving on an expressway stopping may be difficult so destination designation while in motion may be less risky There is great concern as to what a driver can do while in motion Zwahlen and DeBald 1986 Zwahlen Adams and DeBald 1988 Previous research Several studies in the literature have examined the entry of
21. is a possibility that some systematic errors could have occurred although these are unlikely to have affected the total destination and retrieval times the basis for the data analysis Perhaps the most probable type of error would be the analyst missing errant keypresses which had no effect on the functioning of the Ali Scout For example it would be difficult to tell if a subject accidentally pushed K instead of the zero button two adjacent buttons during coordinate entry because the system provides no feedback if a letter is pushed when only a number is allowed A second type of error would result from the analyst being unable to distinguish whether the shift key was pushed twice in a row or not at all This would be an issue for some subjects who occasionally did not push the buttons hard enough for them to register Because the method for obtaining the entry and retrieval times consisted only of measuring the time from the first to last keypress neither the subject s thinking time the time between receiving the card and the first keypress nor confirmation time the time following the final keypress were captured This method was used because the exact time a card was handed to each subject could not be established and the exact time the subject finished could not be determined because the subjects were inconsistent in their use of the found key An unfortunate side effect of this method is that zero times were sometimes obtained for one
22. is significant p 0 0012 and the effect plot in Figure 22 shows that it is the middle aged and older subjects for whom the first block was significantly slower than the other two Young subjects did not experience such a significant drop in times Thus a bit more practice on the entry task than occurred in this experiment was needed particularly for the benefit of the middle aged and older subjects 140 O Block 1 U Block 2 9 Block 3 S 80 Young Middle Older Age Figure 22 Interaction between age and block number The error bars represent 95 percent confidence intervals 36 The general pattern of errors was similar to the time data decreasing slightly with practice though the effect was not statistically significant See Figure 23 As mentioned above a similar analysis for the retrieval task would reveal nothing because of the small number of errors 1 2 5 card total errors Block Figure 23 Total number of errors for the destination entry task as a function of practice block number The error bars represent 95 percent confidence intervals How did performance time and errors vary as a function of driver age and sex For destination entry there were enough mistakes overall to allow analysis The other available measures of performance completion time for every entry or retrieval task are the primary performance measures used in the analysis Figure 24 below shows the individual me
23. of the destination retrievals WOODSIDE HOSP because only a single keypress was necessary if the character search method was used Overview of the entry and retrieval data To recap of the 30 name cards shown to each of the 36 subjects 15 were destinations to be entered into the Ali Scout and the remaining 15 were to be retrieved from it The cards were divided into three sets referred to in this report as A B and C The cards in each set were always shown in the same order The five retrieval cards were shown first followed by the five entry cards Each set was presented under a different experimental condition real device with dusk lighting real device with night lighting or the simulated device at dusk but the card set condition pairings were rotated so that each set was seen in each condition exactly one third of the time Times were recorded from the first keypress to the keypress completing the entry This does not include use of the found key since it was used irregularly nor the 1 to 2 seconds subjects were observed to spend planning their response There are no times missing from the data set so there are 540 236 x 15 times for destination entry and 540 times for destination retrieval The histograms of these times Figure 19 reveal a skew to the right the extreme outliers are all very long rather than very short times It is for this reason that median rather than mean times give a more accurate sense of typical t
24. two groups The number of errors for the retrieval task is too small to analyze Destination Entry s 5 85 6 39 ee re caters anna nn ENTRY N O O a Ch 9 D Entry Mistakes out of 15 Oo NH genen mel armo rane rt Young Middle Older Age Figure 26 Destination entry errors per subject by age and sex 39 What subject factors other than age and sex influenced performance in this experiment Of the biographical questions comfort with maps prior use of a touchscreen frequency of computer usage typing proficiency only computer usage was significantly related to performance in this experiment As a reminder usage was measured using a 5 point scale 1 2 less than once a month 3 a few times a month 4 a few times a week and 5 every day The mean response was 3 4 so the subjects were partitioned into a below average responding 1 2 or 3 and an above average 4 or 5 group The two level factor thus created has a very strong effect on times p values were smaller than 0 0001 in a one way ANOVA The means for the two levels of this computer usage factor are given in Table 19 The interaction with the age group factor seen in Figure 27 is very revealing even though it is significant only in the case of destination entry Note that the young below average group only contained one subject and the older ab
25. up the coordinates on a map current location indicate the current location is the destination points of interest find the coordinates in a list of places and address range find the coordinates in a list of street addresses A more detailed description of these four methods appears in the Test Plan section of this report Subjects reported that the current location and points of interest methods were easy to use and that the address ranges and map methods were difficult to use Both surveys indicated the same results About 70 percent of the trips involved a destination already in memory Table 6 lists the frequencies Subjects thought this feature was easy to use about one half indicated that it was very easy to use Table 6 Frequency of destinations Reported in Driver Logs y Destination Frequency Frequency Home 2 Work Shopping Entertainment lt 1 e g movie School lt 1 Friend s Relative s home 1 Restaurant Bar 7 Recreation e g golf course 2 Some 53 percent of the subjects indicated the keyboard was easy to learn while 29 percent thought it was difficult to learn The remaining 22 percent were undecided or did not respond In terms of ease of use 49 percent felt it was easy to use 38 percent felt is was difficult and 11 percent felt it was neither easy nor difficult In terms of 11 reliability 25 percent thought the keyboard did not function properly Finally with regard to their ove
26. variety of strategies to support the lower panel and keep it from moving Providing a more positive stop or latches to hold the panel open would help 53 How accurate were subjects in looking up coordinates in the manual Subjects made numerous errors in looking up coordinates in the manual For finding intersections and street addresses the error rate was 20 to 30 percent For points of interest there were no errors This finding does not speak well for the effectiveness of the documentation This finding also raises questions about the viability of a system that requires use of longitude and latitude for on road navigation While those coordinates may work well for users of air sea and off road navigation systems drivers rely on street or landmark related information for guidance While the point of interest error data suggest it is possible to design a reliable on road navigation interface that requires coordinates the other two data sets suggest that use of longitude and latitude are not desired Users are unlikely to find a navigation system acceptable if that system directs them to the wrong destination 25 percent of the time because they selected the wrong location While some of those errors are small actual misdirection rates should be even greater as entry errors are likely when the information entered longitude and latitude is not meaningful to users How did performance with the Ali Scout interface compare with othe
27. 0 116 ENTRY 100 90 HE Ni 80 Night 70 Dusk 60 50 5 Mean Entry Time 30 Young Middle Older Age Figure 28 The significant illumination age interaction for destination entry Error bars represent 95 percent confidence intervals As stated earlier there were not enough errors committed during the retrieval task to analyze dusk night differences Analysis ofthe errors for the entry task does not reveal any significant difference between the dusk and night conditions Each subject saw five cards in the dusk condition and five in the light condition This means that there were 180 cards seen in each condition There were 14 total errors made in the night condition and 12 made in the dusk condition averages per subject of 0 389 and 0 333 respectively This is not a significant difference according to one way ANOVA p 0 74 Furthermore there is no significant interaction between the lighting condition and either age or sex as far as the number of errors committed is concemed Were the times and errors the same for real and simulated interfaces The analysis in this section is very similar to that ofthe previous section Whereas the previous section considered only the dusk lighting and night lighting conditions for the real Ali Scout and disregarded the simulated data this section will consider all three levels of this condition factor with the primary emphasis on comparing the simulated level with the other two le
28. 3 12 70 10 69 10 48 Entry Times s Place Name NICKS PLACE Q GAS HELENS KITCHEN YAW GALLERY GOODYEAR FARMER JACK TACO LOCO FIRST OF AM JACOBSONS CHEVRON LARK REST UNICORN GRILL KROGERS QWIK STOP TUFFY AUTO Overall Median Maximum 60 16 47 67 55 25 54 19 39 45 67 55 92 50 54 98 48 28 44 00 61 74 55 59 43 96 54 43 45 67 51 48 209 95 275 88 203 22 252 55 135 73 351 30 151 70 159 20 185 70 100 63 243 73 238 53 101 52 436 45 131 72 436 45 Note Retrieval always begins with the place name AT THE START the first alphabetic entry shown Mean 77 00 69 55 72 85 67 82 51 05 84 31 57 98 64 55 60 30 46 96 72 03 74 17 47 28 73 18 51 12 64 68 20 90 18 0 85 16 80 O 14 75 B 70 i Overall g 65 Mean 10 9 8 8 O 60 Overall 55 8 Overall 6 Median 50 Median 4 5 45 5 2 40 0 Medians Means Medians Means Destination Retrieval n 15 Destination Entry n215 Figure 20 Distribution of entry and retrieval of the means and medians times across locations in seconds Table 15 shows that a median time for the retrieval task from first keypress to final keypress was a bit more than 6 seconds yet the maximum time exceeded two minutes Similarly the median entry time is around 51 5 seconds but could be as long as seven minutes The times given above for entry are probably closer to reality
29. Current Selected Destination Scroll Character New Search Search Destination Note For Volume and Brightness press EC C3 to adjust amp to accept Enter alphanumeric Enter new Search Destination name Volume Destination O Y Scroll upidown Dest hst A Special press n 52 Dest list qu to accept and end Y Brightness daymight quU accep Press Ed 1 E Compass mode to accept 2 en i a Variable Params tor future use and coordinates Demo pass cursor arrows to edit d 9 to save and end or press E press O S to save and end E Demo mode x or O v oti a O Y Once ition to accept and end to save current Parking info for future use inicio coordinates and end D Route type lor tuture use O Street enter street name Zip enter zip code Town enter towrvcity 3 Phone enter phone number O Country enter country use E to scroll ist to save and ena Figure 16 Ali Scout Menu Structure The original project plan called for evaluating both real and simulated Display Units under simulated dusk and night conditions However pilot tests showed no differences due to illumination for the simulated unit so only the simulated dusk condition was explored in the main experiment Varying light levels were explored for the real display In the simulation all items to be read LCD text and key labels were highly legible and beca
30. ER m me 1121314 5 6 7 8 9 0 EJ R LO Figure 7 Qwerty style keypad interface design Inthe phone pad interface subjects were shown numeric keys on which multiple letters appeared See Figure 8 So for example to enter ELUM subjects typed 3586 Because multiple text entries could match any string of digits e g FLUN would also match a scrollable list of alternatives was then presented for final selection See Figure 9 gt o Figure 8 Phone style keypad entry method PRISES Select Street COX CRANBROOK CREST CRESTON CRESTWOOD CROWN CULLOUGH CULVER CUMBERLAND Figure 9 Selection screen for double press and Qwerty name selection The fourth interface style resembled that in the Zexel navigation system currently found in the Pathmaster sold by Rockwell Figure 10 To move the scrolling list forward or backward one character in the alphabet the A and Z keys are pressed To move forward or backward one entry in the list the up and down arrows are used Select CITY A JASPER KINGSLAND KNIGHTSTOWN LAWRENCE LEWISVILLE LINDEN Figure 10 Scrolling List Interface Design Eight destinations were entered for each interface design while parked or while driving a very simple simulator Destination entry times were 43 seconds for the phonepad 10 44 seconds for the Qwerty 55 seconds
31. Fewer errors were made for the intersection of Auburn and Ryan Ryan follows the grid line closely enough for the subject to obtain the correct value no matter which method is used Auburn dipped down very suddenly at one point making it more likely that the subject would trace his finger along the grid line rather than the street These errors tended to be small 1 to 2 tenths of a mile and since the Ali Scout unit goes into autonomous mode before the end of the trip drivers may still be able to find the destination Table 21 Coordinate Finding Errors Task Destination Description Errors Intersections On Map 1 Giddings 8 Walton On Map 2 Auburn amp Ryan Street Addresses 2300 Coolidge Hwy 7 2 WERE 4429 Sashabaw Ad 2 56 o 0 _ 0 0 Points of Interest Troy Union School Glen Oaks Golf Course Nicholas George Theater 0 0 Address range coordinate errors which accounted for 23 percent of the trials were most likely due to the subject accidentally choosing the wrong range or being unable to find the correct range because ranges were listed in numerical order by town for each street rather than simply numerical order for each street In addition one of the difficulties in finding the correct coordinates for 450 12 Mile Rd W was that it was often confused with 12 Mile Road which was listed in the preceding column of the manual 47 Finally the absence of errors in finding coordinates for points o
32. HI MNO e Pols ee Modfied Modal Method JKL WXY TUV DEF GHI E HEB Figure 2 Key Sequences to Enter KWV Using Various Methods Fifty adults were timed as they entered 24 six letter strings Table 3 shows the results There were no statistically significant differences in the entry times though the error differences were significant In part this was because the repeat key method minimized finger movements and the probability of striking the wrong key Interestingly GOMS model Card Moran and Newell 1983 predictions of entry times using the Slowman assumption matched the actual times fairly well except that the rank orders of the Modified Modal and Top Row methods were reversed This provides an indication that GOMS models could be useful for keypad entry tasks Table 3 Entry Times and Errors for Vanous Methods Method GOMS s Hepeat ke 11 96 Modified modal 13 72 Top row 80 1350 13 50 Same row 13 78 Modal position 14 58 Several studies have reported times for destination entry and other tasks examined during the development of interfaces for operational field tests As part of the ADVANCE project Loring and Wiklund 1990a describe and evaluate three prototypes for keypads allowing destination and other data entry There were 12 subjects ranging in age from 19 to 68 meanz41 Figures 3 4 and 5 show the designs evaluated Twelve subjects performed sample tasks using low fidelity SuperCard p
33. IZZA 0 5 WOODSIDE HOSP 0 2 BIR THEATER 0 6 MONGOLIAN BBQ 0 2 Overall Rate 1 5 Although the errors in the retrieval task were too rare to be of any use in analysis the specific types of errors committed deserve explanation The high number of errors in retrieving BIR LIBRARY Was due to four subjects who scrolled until they reached BIR ART GALLERY and then stopped The error in BIR ICE ARENA was due to the subject pressing the right arrow key when a space was appropriate This sent the Ali Scout into entry mode an action that can only be reversed by pressing cancel and starting over The remaining three errors were due to spelling mistakes BIR LIBRARY was typed as BUR LIBRARY MONTERREY REST as MONTERRY REST and VANDENBURG SCH as UANDENBERG SCH his last error which was due to typing U instead of shift U tended to be difficult for subjects to catch because U and V are difficult to distinguish on the Ali Scout s display Did performance change with practice To examine learning entry and retrieval will be treated separately here as they will be throughout the analysis Each subject saw 15 of each type of card and the question of interest is whether on average subjects did better on the later cards than the earlier ones However the trials numbered 1 to 15 should not be considered as separate levels of a single factor in ANOVA since the cards were always ordered the same way in each group of 5 For example the SAKURA BANK card
34. OE OO MO RN RTCA LOSES ae SOLE NA ARUN LR OPAC AREER S P EA III RR ETD SCAR M NAA AME SE NE UR RnR TOD 20 10 02 21 03 15 08 06 00 24 20 25 Unicorn Grill 25 22 22 22 04 29 BETTER Change gt NN IE M Go Time Figure 18 VCRTimelt 1 0 User Interface Second VCRTimelt can automatically cue the tape to any event that has already been marked thus allowing the user to easily find and check events of interest a feature neither A C T nor EventRecorder support Third data collected with VCRTimelt can be edited at any time With A C T the analyst cannot even see the output file until the analysis is complete Finally VCRTimelt records time data to the nearest 1 30th of a second while EventRecorder only does so to the nearest second After the initial pass at the data two blocks of trials 325 total keypresses for one subject were reanalyzed to verify that the data reduction procedure was repeatable There were only two discrepancies in the number of keypresses between the original and repeated analysis and the time difference between the corresponding keypresses never exceeded 0 3 seconds except for a single keypress where the difference was 0 5 seconds The vast majority of times were either equal or different by 0 1 seconds 30 Analyst delays were not perceptible since the analyst could predict when events would occur In spite of these checks there
35. Technical Report UMTRI 96 30 December 1996 Destination Entry and Retrieval with the Ali Scout Navigation System Aaron Steinfeld Daniel Manes Paul Green and David Hunter HUMAN FACTORS Technical Report Documentation Page 1 Report No 2 Government Accession No 3 Recipient s Catalog No m UMTRI 96 30 4 Title and Subtitle 5 Report Date Destination Entry and Retrieval with the December 1996 Ali Scout Navigation System 6 Performing Organization Code account 032923 7 Author s B Performing Organization Report No Aaron Steinfeld Daniel Manes Paul Green UMTRI 96 30 and David Hunter 9 Performing Organization Name and Address 10 Work Unit no TRAIS The University of Michigan Transportation Research Institute UMTHI 2901 Baxter Rd Ann Arbor Michigan 48109 2150 12 Sponsoring Agency Name and Address 13 Type of Report and Period Covered Road Commission of Oakland County RCOC 10 94 12 96 31001 Lahser Beverly Hills MI 48025 USA 15 Supplementary Notes This research was funded by RCOC and the Federal Highway Administration After training 36 drivers retrieved and entered a total of 20 destinations using an Ali Scout navigation computer and 10 destinations using a touchscreen simulation while sitting in a vehicle mockup Retrieval involved keying in part of the destination name scrolling through a list of names or some combination of those methods Entry required keying i
36. The difference is even more than double when mean times are compared instead of median times because the older age group tended to include more extreme outliers with the result that the mean time for the older group is more skewed than for the other two groups Differences in destination entry error rates were even larger For young and middle aged subjects the error rate was approximately 5 percent For older subjects the error rate was 20 percent a factor of four difference What subject factors other than age and sex influenced performance in this experiment Of the factors examined comfort with maps prior touchscreen use frequency of computer use typing proficiency only computer usage was correlated with performance This effect is not simply due to age differences even though there was on average more computer experience represented in the sample of young subjects than in the older group the effect of computer experience was also seen within age groups particularly the middle age and older groups How did time and errors vary as a function of ambient illumination Mean retrieval times were approximately 8 seconds under dusk conditions 9 seconds at night a difference about 10 percent that was not statistically significant Means are reported here instead of medians because ANOVA is based on means and this section and the next concern only comparisons using ANOVA The medians may be found in the results Mean entry times w
37. al entry method and the interface implementation While what people can and cannot do safely while driving was not explored here lumping all destination related activities together does not make sense Further because the time required depends on the method how a particular task is carried out needs to be considered These differences in method are described in the follow on report in greater detail The research presented here provides engineers and designers with methods to evaluate navigation interface usability normative data on the retrieval and entry of destinations estimates of the individual differences and a list of problems associated with a contemporary navigation product This information should be useful to engineers and designers involved in future navigation systems 96 REFERENCES Archer N P and Yuan Y 1995 Comparing Telephone Computer Interface Designs Are Software Simulations as Good as Hardware Prototypes International Journal of Human Computer Studies 42 169 184 Beevis D and St Denis G 1992 Rapid Prototyping and the Human Factors Engineering Process Applied Ergonomics 23 3 155 160 Berger C Walton C and Wurman P 1993 The Event Recorder version 5 2 Ann Arbor MI The University of Michigan Office of Instructional Technology Card S K Moran T P and Newell A 1983 The Psychology of Human Computer Interaction Hillsdale NJ Lawrence Erlbaum Associates Coleman M
38. an times for both destination entry and destination retrieval Each age sex category e g older females contains six subjects Even though medians give more representative typical times means are plotted in Figure 24 because the ANOVA tests seen later are based on means Notice the considerable amount of scatter in the middle aged men and the older subjects in general 37 RETRIEVAL ENTRY 2 F M F M F M F M F M F M Young Middle Older Young Middle Older Figure 24 Mean times by age and sex ANOVA models of destination entry and retrieval times included sex age sex by age interactions and the effect of subject nested within sex and age For retrieval time the effects of sex p 0 0055 age p 0 0001 sex age p 0 0388 and subject p 0 0001 were all statistically significant For entry time sex p 0 0001 age p 0 0001 and subject p 0 0001 were all significant ANOVA tables for these analyses appear in Appendix K Figure 25 shows the interaction plots for age and sex Post hoc tests reveal that every age category is significantly different from every other for each task Therefore it is possible to generalize that women responded more quickly than men and that times increase with age RETRIEVAL ENTRY interaction significant interaction not significant 20 7 100 I 2 a rn e 807 514 5 70 8123 Eo 10 TRE S 50 E 50 4 56 M 4 Middle Older Age Young Mid
39. both displays and the card holder appear in Appendix B 19 Figure 15 Location of Displays The Ali Scout unit can store up to 80 destinations To retrieve a destination subjects could use one of three strategies type in the name of the destination which appeared when the characters entered uniquely matched the beginning of that name Character Search scroll through the list of names to the entry desired Scroll Search or type the first character or two and then scroll the rest of the way Hybrid Method Table 9 shows the keystrokes necessary to obtain SEARS as a destination Assuming the subject had not memorized the data base there were 21 locations in the main list the minimum character strokes required using either method would be 4 The first down arrow is required to enter the scrolling function Table 9 Methods for Character Searching Location List Alphanumeric Method Hybrid Method ntered Displayed Entered Displayed AT THE START down arrow SAKURA BANK down arrow SEAFOOD BAY down arrow SEARS SAKURA BANK SEAFOOD BAY SEARS VANDENBURG SCH moot 20 UU dr ciem pe ep On occasion fewer keystrokes may be required by entering two characters prior to scrolling Also if the subject has extensive knowledge of the list he or she can enter a character that will place them at the other end and scroll up e g T and two up arrows for
40. could only appear on the first sixth or eleventh trial so even an existing practice effect might not be evident if the intervening trials consisted of much harder or much easier cards Thus the mean times for each block of 5 trials will be considered trials 1 to 5 6 to 10 and 11 to 15 This gives 3 levels instead of 15 and the cards in each of the 3 levels are exactly the same so the effect observed is independent of differences in reaction times for individual cards 35 Figure 21 shows the three level block factor described above Based on an ANOVA model with one main effect there was a significant learning effect for the destination entry task p 0 0001 but not for the destination retrieval task p 0 062 Specifically for entry block 2 and block 3 both have significantly shorter entry times than block 1 but blocks 2 and 3 are not significantly different from each other 16 90 oe RETRIEVAL ENTRY not significant 8 significant 2 14 gt 80 13 75 S 12 2 11 EU 2 amp 10 oos c 9 c 60 8 2 55 7 50 Block1 Block2 Block 3 Block1 Block2 Block 3 Figure 21 Effect of practice block number on task time The error bars are 95 percent confidence limits These results suggest that the subjects received enough practice on the retrieval task prior to the first block since no significant improvement occurred after that point Exploring the block effect a bit further the age block interaction
41. dle Age Figure 25 Age sex interaction plots for destination retrieval and destination entry 38 Tables 17 and 18 give median and mean values for each age group and each gender group To summarize median entry times for older subjects were double those of younger subjects with middle aged subjects midway between them for both entry and retrieval tasks Men typically took 10 to 20 percent longer than women Table 17 Median and mean times for each age group Destination Entry s Age Group Median Ratioto Mean Young Destination Retrieval s Age Group Median Ratio to Young 4 27 1 00 5 71 Young 37 75 1 00 Middle aged 7 27 1 70 9 58 Middle aged 52 31 1 39 Older 8 59 2 01 16 15 Older 75 52 2 00 Table 18 Median and mean times for men and women Destination Retrieval s Gender Median Ratio to Mean Gender Median Ratio Mean Female Female 47 68 1 00 58 95 1 00 1 09 Figure 26 shows the interaction between age and sex in determining the number of errors made on all fifteen entry cards An ANOVA with number of mistakes as the dependent variable concludes that the age effect is significant p 0 0001 but the sex effect p 0 28 and the age sex interaction p 0 79 are not As the figure suggests a post hoc test confirms that the older subjects made significantly more errors than either the young or the middle aged subjects but there was no difference between the latter
42. e desired result Problem 3 The space function does not have its own key but rather appears on the same key as the L This requires the user to press two keys for each space one of the most frequently used characters Also the symbol used to represent a space essentially the bottom half of a square was a source of difficulty for nearly all of the subjects Problem 4 Subjects thinking of the device as a typewriter often entered a space character instead of the right arrow to move the cursor to the right causing characters to be erased instead of skipping past them During the initial learning period very few of the subjects were able to figure out how to enter a space even though they had just watched a video which mentioned the method Also many of the older subjects and some of the middle aged subjects forgot how to enter a space by the time the main part of the experiment began while the younger subjects tended to remember but often hesitated Problem 5 Many subjects were often confused about how to advance to the next step when entering destinations This was seen most frequently when subjects tried to advance from entering a name to entering coordinates This was not due to problems with text legibility When asked to rate the display contrast 12not enough 2 just right 3 too much the ratings indicated contrast was close to adequate mean rating 1 8 for the real unit 1 9 for the simulation Problem 6 Subjects oft
43. e distinguishable color code Also the graphic used to represent shift was easily confused with the various arrow cursor keys provided Additional changes to the shift graphic e g filling it in or indicating that it is to be used with the right side character of each letter pair may help overcome selection of the wrong key 51 Problem 3 The space key was too difficult to access Use ofthe space key was quite common in destination entry However two keystrokes were required to enter a space Further the typewriter model a large bar below all other keys was not supported by the interface Providing a space bar that fits the accepted user model would require redesigning the keyboard Less costly improvements include making the space a separate button and using a more easily understood label for it Problem 4 The space key was misused Some subjects thought the space key would behave as it does on a typewriter They did not understand that spacing over would erase not merely move the cursor It is not apparent how to solve this problem Problem 5 How to change fields was unclear This occurred after a name was entered and the next step was to enter coordinates One option might be a graphic on the keypad showing how the cursor keys could be used to change fields Problem 6 The two sets of arrow keys were confused with each other The open arrow and closed arrow keys look too similar and their functional difference
44. e unit When hand you a card enter the information listed on it When you have finished entering the information into the unit please press the FOUND key and put the card in the envelope Do you have any questions Please tum on the system e Run the entry series e Have the subject hand you the cards from the envelope e Have the subject close the door and turn off the Ali Scout e Sort the cards e Turn on the room lights If necessary have the subject step out of the car 75 e Clear all unwanted destinations from the Ali Scout and select AT THE START as the current destination or quit the simulator gt Retrieval Lists Nicks Place Q Gas Helens Kitchen Yaw Gallery Goodyear Farmer Jack Taco Loco First of Am Jacobsons Lark Rest Unicorn Grill Licht Park Nordstrom Discafe Oakland Mall Olive Garden 0831732W 0831654W 0831649W 0831303W 0830508W 0830937W 0831932W 0832459W 0830905W 0831707W 0832250W 0830848W 0830506W 083235 3W 0831642W 0830731W 0831211 0831732 08306384 0830828W Royal Oak Deli Sears Bir Art Gallry Palace of AH 423814N 424038N 424002N 423302N 423006N 423159N 423750N 424307N 424054N 423032N 423236N 422919N 422923N 423936N 42273 7N 423051N 422959N 423814N 423203N 423047 e Rewind and eject the tape e Turn off the power strip of the audio cart e Turn off the touchscreen car lights overhead projector and PC projector 76 Entr
45. ed How can the Ali Scout be improved GOMS predictions vs Can predictions be substituted for usability tests where age real and simulated is an issue not examined by Detweiler and touched upon interfaces by Paelke This could reduce the number of costly usability tests needed during development Looking up coordinates Ali Scout is unique in that it uses longitude and latitude Does that create problems To speed dissemination of the results findings concerning the GOMS predictions are covered in a subsequent report Manes Green and Hunter 1996b 15 16 TEST PLAN Test participants There were 36 subjects in the experiment 12 young 18 30 12 middle aged 40 55 12 older over 65 The young and old groups represent the population extremes The middle group comprises the most likely buyers Within each age group there was an equal number of men and women All subjects were licensed drivers 1 000 to 40 000 miles per year mean of 13 000 The sample was well educated All but three had at least some college and 13 had at least one college degree Vision ranged from 20 13 to 20 22 corrected in the young group 20 13 to 20 40 in the middle aged group and 20 15 to 20 40 in the older group Only one subject had previous experience with a navigation system and 20 of the 36 had used a touchscreen Computer use was moderate on average 3 4 where 3 a few times a month 4 a few times a week and subjects reported they were m
46. ed For alphanumeric entries the associated key was the equivalent lower case character For special functions upper case characters were entered such as D for diamond C for cancel and Y for shift Errors introduced from analysts keypresses to obtain times were extremely small See Manes Green and Hunter 1996b VCHTimelt was developed instead of using existing time study programs such as Activity Catalog Tool A C T Segal and Andre 1993 and EventRecorder Berger Walton and Wurman 1993 for four reasons First because VCRTimelt is directly linked to the VCR the user can vary the speed and even reverse the direction of the tape at any point during the analysis maximizing the speed of the analysis and allowing for easy error correction A C T and EventRecorder both require a constant playback speed throughout the analysis For both A C T and EventRecorder pausing the tape during the analysis is difficult because the user must press a key on the Keyboard and a button on the VCR at the same time Pausing tends to occur when the analyst is interrupted or needs a break 29 i LIA ne ee d p A E m M IA AMAIA PERLAS EEE LAL s PAR APRA AAA ARENAS LEE ELL NAILS SS BELT LARA ALI ALES Pit LLL ELL LLL OE R
47. ed for a response but without effect as the key had not been pressed hard enough 46 Problem 11 Finally the hinged panel on which the bottom two rows of the keyboard were placed felt flimsy to many of the subjects causing them to use their thumb to support it from underneath as they typed How accurate were subjects in looking up coordinates in the manual A key difference between Ali Scout and other navigation systems is that Ali Scout uses longitude and latitude not street addresses or intersections to identity destinations As a consequence users are required to refer to a manual to obtain the coordinates for each destination they wish to enter into the Ali Scout The manual is divided into three sections maps which can be used to look up the coordinates of street intersections address ranges and points of interest Thus subjects were asked to look up three sets of coordinates from each of the three sections The associated error data from all coordinate finding tasks are presented in Table 21 An error was recorded whenever a subject looked up and reported coordinates that disagreed with the correct coordinates When searching for intersections on a map a task for which the error rate was almost 30 percent several subjects obtained incorrect coordinates from the maps because they traced their finger along the path of the road rather than the appropriate grid line and thus obtained a number that was off by several units
48. en used the wrong set of arrows when trying to move the cursor around sometimes necessitating a large amount of retyping or error correction Interestingly when asked in the post test questionnaire if the real and simulated interfaces were logical all but one subject responded yes Problem 7 The cancel feature required the user to push two buttons simultaneously when the pane was closed which sometimes required subjects to push harder than felt comfortable Very few subjects were confident at first that the system actually intended them to push two buttons simultaneously 45 Problem 8 The symbols used to represent cancel and enter an X d circle and a diamond with a vertical line inside it respectively were unsuccessful in conveying meaning Problem 9 The size and spacing of the keys were criticized by many of the subjects for being inadequate and prior to the experiment many subjects expressed doubt that they would be able to use such a cramped keyboard In the subjective evaluation the mean rating of key size was 1 3 1 too small 2 just right 3 too big for the real Ali Scout unit 1 5 for key spacing For the simulated Ali Scout unit the ratings were 1 6 and 1 5 There were no age related differences apparent under statistical evaluation However a quick examination of the results suggested that there may have been differences had there been a higher resolution for the subjective scale The significan
49. ere 54 seconds at dusk 66 seconds at night 22 percent greater a statistically significant difference Particularly noteworthy were problems older drivers had in the night condition For them retrieval times were 70 seconds at dusk 100 seconds at night much greater than other groups in the sample While part of the difference may be due to experimental artifacts nonuniform illumination of the device with particular difficulty in seeing the top row they do not completely account for the age differences This finding emphasizes the need to examine the legibility of self illuminated ITS products under nighttime conditions using the least capable subjects namely older drivers 50 There were no statistically significant dusk night differences in terms of errors Were the times and errors the same for real and simulated interfaces No The mean entry times were 54 and 66 seconds for the dusk and night conditions with the real interface and 74 seconds for the simulated interface The associated medians were 44 50 and 58 seconds Thus tasks using the simulated interfaces took roughly 25 percent longer to complete For entry times the comparable values were 8 9 and 15 seconds with medians of 5 6 and 8 seconds In that case the difference is much larger nearly a factor of 2 A major reason for these differences is how older drivers responded at night Also differing was the number of errors 12 for dusk 14 for night and 28 for
50. experienced subjects For both groups of subjects the relationship between entry time and key size was linear More to the point of this project Paelke 1993 see also Paelke and Green 1993 describes an experiment comparing four different interfaces for destination entry These interfaces were designed to capture the variety of entry themes present in contemporary products Sixteen drivers 8 young 8 old served as subjects Interfaces were presented on a touchscreen CRT of which a 5 inch 12 7 centimeters diagonal section the size of a typical navigation interface was visible For the double press method a matrix of 10 touch areas appeared with up to four characters per area EFGH in Figure 6 Touching one of the multicharacter buttons caused touch areas to appear one for each character Touching the desired character E F G or H caused it to be selected This design was similar to that used in the TravTek interface a field test conducted in Orlando Florida This design overcame limitations of the resolution of the touch interface Enter first 4 letters of _ STREET no Figure 6 Double press interface design In the Qwerty interface the arrangement of touch areas resembled that of a typical keyboard While the arrangement was familiar to typists the keys were smaller than is typical for keyboards The arrangement was used for the ADVANCE interface examined in field trials in Chicago Enter address NUMB
51. f interest is most likely due to the manual listing them in alphabetical order and the destinations chosen for this experiment did not require the subject to choose between more than one address Although no statistical analysis was performed to determine times for the coordinate finding task it is estimated that subjects took from 40 seconds to over 1 minute for maps 30 to 45 seconds for address ranges and 15 to 30 seconds for points of interest These experimenter estimates correspond with the error data and the reasons for the time differences should follow those outlined previously In addition the times for the map would tend to be longer because the streets are not indexed forcing the subject to search the whole map to find the streets Many of the subjects took a long time to find 450 12 Mile Rd W because they did not know if numbered streets were listed by their spelling or at the beginning or end of the section 48 CONCLUSIONS AND DISCUSSION How can the analysis protocol be improved A major development of this project was the creation of a tool for controlling frame accurate VCRs and using that tool to obtain reasonably accurate keystroke time data The analysis method proved to be reliable While tools for this purpose are commercially available they tend to be quite expensive and are not perfectly matched to the analysis tasks at hand This project would have taken several months less time and cost considerably less if
52. for the scrolling list and 76 seconds for the double press interface The overall differences in entry methods were highly significant Older drivers entry times were 21 percent longer than those of younger drivers Having people drive the simulator while entering destinations increased the entry times by 28 percent over the baseline parked condition The standard deviation of lane position increased from 6 inches baseline to about 11 inches dual task The order of the error data from best to worst was identical to that of the entry times Also identical was the order of the GOMS model predictions of performance Thus this experiment demonstrated the utility of GOMS estimates for predicting driver performance with realistic navigation interfaces Eby Streff Wallace Kostyniuk Hopp and Underwood 1996 described a pilot test concerning user perceptions and use of the Ali Scout interface A total of 62 drivers had Ali Scout units installed in their vehicles for two months These were mostly professionals at General Motors and Chrysler almost 60 percent had a household income of 100 000 or more Of them 45 completed the first survey one week after participation and 36 completed the second two months after participation Drivers also completed a daily log of their trips With regard to entering destinations the following was learned In terms of frequency of use the rank order of destination designation methods was map method look
53. g to cleanly delineate when subjects are receiving instructions on how to complete a task and when they are thinking about how to complete it Some thought needs to be given to what the basis for comparison should be The times just for keying in this experiment were mean times of 10 seconds for retrieval and 65 seconds for data entry An additional 1 2 seconds were required prior to each task for thinking A major addition unique to this interface is time to look up the 54 coordinates 40 to 60 seconds for intersections on maps 30 to 45 seconds for address ranges and 15 to 30 seconds for points of interest Aggregating times together this suggests that about 11 seconds were required for destination retrieval but 90 to 120 for destination entry Three other experiments described in the literature contain data that may be used for comparison Loring and Wiklund 1990b report several tasks that have similarities to those explored here Recalling a trip with a particular name took 33 seconds a task that may be similar to the retrieval task explored here which only took one third of that time Deleting a destination similar to retrieval but with an added step took 68 seconds Telling the device to plan a route to a destination and saving the trip plan just driven took 85 and 134 seconds respectively Thus compared with an early version of the interface used for the ADVANCE project the Ali Scout interface times were considerably less
54. h a navigation system yes no How comfortable are you using maps very moderately neutral moderately Very comfortable comfortable uncomfortable uncomfortable Have you ever used a touchscreen yes How often do you use a cornputer never less than a few times a few times every once a month a month a week day How comfortable are you typing on a standard typewriter or computer keyboard very moderately neutral moderately very comfortable comfortable uncomfortable uncomfortable TITMUS VISION Landolt Rings 1 2 3 4 5 6 7 8 9 10 1 12 13 14 T R R L T B L R L B R B T R 20 200 20 100 20 70 20 50 20 40 20 35 20 30 20 25 20 22 20 20 20 18 20 17 20 15 20 13 67 68 APPENDIX E CONSENT FORM Evaluation of Destination Entry and Retrieval Participant Consent Form In cars of the future you may have an in car navigation system which would tell you how to reach destinations To use it you ll need to tell the system where you want to go We are looking at how to enter and retrieve these destinations for a navigation system Responses from typical drivers such as you will help improve this system While sitting in a vehicle mockup you will respond to a simulated or a real navigation unit By pressing buttons destinations can be entered or retrieved A computer will record how long it takes to use the system With your permission we will videotape you We will not record your face at any time We will not release any
55. h practice Entry times retrieval times and entry errors all decreased with practice but practice effects were statistically significant only for entry times From the first to the third practice block times decreased by about 20 percent Thus by some criteria subjects were moderately proficient at retrieval related tasks before the first block How did time and errors vary as a function of driver age and sex There were large significant differences due to driver age and sex For retrieval median times were 5 9 seconds for women and 6 4 seconds for men with means of 9 49 and 12 seconds respectively These suggest a difference of 9 percent There were too few errors to examine either sex or age differences for retrieval The median entry times were roughly 48 seconds for women 56 seconds for men with means of 59 and 70 seconds respectively Hence men took about 17 percent valued to enter destinations However error rates for the completed entry were about the same Differences due to age were much larger For retrieval the young middle aged older driver median times are 4 3 7 3 and 8 6 seconds with means of 5 7 9 6 and 16 2 seconds Hence the young old times differ by a factor of 2 The median entry times were 38 seconds for young subjects 52 seconds for middle aged subjects and 76 seconds for older drivers with means of 41 61 and 92 seconds respectively This is again a factor of 2 for the young old difference
56. he task that they would devote to real driving Hence the times for destination entry while concurrently driving are underestimates The 76 seconds reported by Paelke and Green for the TravTek like interface may be an overestimate as there were brief instances for which subjects had to wait for the interface simulation to update something that was much less of an issue for the real TravTek interface The closest comparable task in Dingus et al was entering an unfamiliar destination a task that required much more than was simulated by Paelke and Green While it requires many assumptions it appears that the times for destination entry reported by Paelke and Green are equal to or less than those reported here when only keying is considered much less when coordinate lookup is included Thus in spite of these provisos and numerous complaints about the keyboard by subjects and conflict with good human engineering practice destination retrieval was 55 actually quite quick relative to other systems that have been evaluated to date However destination entry was very problematic The difficulty arose from the use of longitude and latitude as an intermediate step and use of a keyboard that was too small and presented logical uncertainties In closing For the Ali Scout interface retrieval times were approximately 10 seconds while entry times were just in excess of one minute When coordinate lookup is included typical times
57. her a touchscreen simulation of the real product is sufficient for usability assessments The simulation takes much less time to construct and is easier to modify facilitating iterative design Several individuals and organizations made important contributions to this effort and their contributions are gratefully acknowledged Amitaabh Malhotra for programming the Ali Scout simulation formerly of UMTHI Patrick Wei Marie Williams for programming the Ali Scout simulation formerly of UMTRI Sara Naylor for testing some initial subjects formerly of UMTRI Finally the authors would like to thank Cale Hodder of Toyota for encouraging the authors to include Japanese style A3 reports the two page summary prior to the Preface in our technical reports vi TABLE OF CONTENTS INTRODUCTION nu ae 1 Why this topic IS Of Interest en 1 PIEVIOUSTOSBAICH unb ee ee MN 1 Research issues exbolofeq usi nee 14 TEST PLAN ses en 17 Tesk pantlieipanis e P 17 Test materials and equipment a 17 Al Scout interface Eee 17 Driving SIMIO eisen ae 22 Miscellaneous equipment eene 24 Test activities and their sequence eee 24 RESULTS ae 29 Data reduction Melli 29 Overview of the entry and retrieval 31 What were typical entry and retrieval times and error rates for destinali
58. identifying information so your responses will remain confidential The experiment takes about 2 hours for which you will be paid 40 There will be opportunities for you to take a break if needed If you have any problems completing this experiment you can withdraw at any time You will be paid regardless have read and understand the information above Print your name Date Sign your name Witness experimenter It is OK to videotape me yes no circle one 69 70 APPENDIX F INSTRUCTIONS TO SUBJECTS BEFORE SUBJECT ARRIVES Tavern e Highlight the current subject on the coordinate and entry retrieval patterns e Putthe coordinate cards in order according to the coordinate pattern e Put the entry retrieval stacks in order according to the entry retrieval pattern e Open a new VHS tape and label it Destination Entry Subject Date e Turn on all audio video equipment e Insert and cue the Ali Scout video e Write all necessary information on the subject forms Sim Lab e Turn on the two computers e Duct tape the Sim lab door so it doesn t lock e Plug in the microphone e Place the road side transparency and glare guard on the overhead Set up the car for either the touchscreen or Ali Scout e Lock the steering wheel Make a copy of the ALISCOUT file and rename the copy DE S R C0X INTRODUCTION My name 15 and will be the person running the study today If you have any problems com
59. im C Dusk A Night B 1 Night A Sim B Dusk C 1 Dusk B Night C Sim A 2 Dusk A Sim C Night B 2 Night C Dusk B Sim A 3 Sim A Dusk B 3 Night B Sim C 3 Dusk C Night A lt lt O fe Night C Dusk A Sim B repeat for 19 36 80 APPENDIX H PRACTICE SHEET Please find the following locations in the order they are shown Press the FOUND button after finding each one BP SERVICE FIRESTONE MAYAS DELI PLUS BANK 24 SUBWAY Please enter the following locations and their coordinates in the order shown Location Coordinates 0832531W 422805N 0830840W 422926N 0830532W 423534N RN a 0830901 A23TLSN 81 82 APPENDIX RETRIEVAL DATABASE Scroll A gt PRINT GALLERY ROYAL OAK DELI SAKURA BANK SEAFOOD BAY SEARS VANDENBURG SCH OODSIDE HOSP N pa EE Note The scroll and min columns are fixed To enter the scroll mode an arrow key must be pressed 83 APPENDIX J DISPLAY UNIT POST TEST QUESTIONNAIRE Please place a check in the appropriate box for each question There is additional room for comments if needed 1 Are the keys the right size foo small just right too large Simulated 16 spacing between the keys rig ght size To
60. imes since the median is not influenced by extreme values The two panels in Figure 19 have different time scales selected so all data points could be displayed The interquartile range IQR is the gap between the 25th and 75th percentiles the first and third quartiles and is a measure of dispersion which is far less sensitive to outliers than the standard deviation However even though medians 31 and IQR are more sensible choices than the traditional mean and standard deviation for giving an overall sense of subject behavior ANOVA makes implicit use of means Departures from normality were not sufficient to rule out the use of ANOVA Thus means will appear throughout this report wherever comparisons must be made among groups of times 120 DESTINATION RETRIEVAL 100 inset detail of irst 40 seconds 80 60 median 6 23 s 40 meanz10 48 10 20 30 40 2 inter quartile range 9 05 en A 0 60 80 100 120 140 Retrieval Times 5 Count n 540 e 60 DESTINATION ENTRY median 51 48 5 mean 64 68 inter quartile rangez39 21 3 na 0 50 100 150 200 250 300 350 400 450 Entry Times s 4 O 540 O 3 N O Count n ob Figure 19 Histograms of retrieval and entry times Subjects made many more uncorrected errors on average on the entry task than on the retrieval task Of the 540 trials there were only 8 uncorrected errors for retrieval a rate of 1
61. ing B A and Wiklund M E 1990a Report on Test of Low Fidelity Prototypes technical report Bedford MA American Institutes for Research Loring B A and Wiklund M E 1990b Report on Baseline Usability Test of Motorola s Prototype 2 technical report Bedford MA American Institutes for Hesearch MacAdam C C Green P A and Reed M P 1993 An Overview of Current UMTRI Driving Simulators UMTRI Research Review July August 24 1 1 8 Manes D Green P and Hunter D 19962 Glance Frequencies to the Ali Scout Navigation System Technical Report UMTRI 96 42 Ann Arbor MI The University of Michigan Transportation Research Institute in preparation Manes D Green P and Hunter D 1996b Prediction of Destination Entry and Retrieval Times Usin MS Technical Report UMTRI 96 37 Ann Arbor MI The University of Michigan Transportation Research Institute in preparation Marics 1990 How Do You Enter D Anzi Quist Using a Telephone Keypad Proceedings of the Human F rs Soci 4th Annual Meeting Santa Monica CA Human Factors Society 208 211 Paelke G M 1993 A Comparison of Route Guidance Destination Entry Methods Proceedings of the Human Factors and Ergonomics Society 37th Annual Meeting 1993 Santa Monica CA The Human Factors and Ergonomics Society 569 573 58 Paelke G and Green P 1993 Entry of Destinations into Route Guidance Systems Human Factors
62. is unclear The open arrow keys controlled the cursor allowing movement one character up down left or right The solid arrow keys also on the front panel were for scrolling up and down in the stored list of destinations and for accessing special menus e g parking information and route type One possible way to solve this problem would be to enclose each set of keys with a border and provide a label indicating their function Problem 7 Having to push two keys simultaneously to cancel was not intuitive Although requiring simultaneous keypresses can prevent accidental activation of unwanted functions it is an unfamiliar concept to many and the existence of such a feature may never be comprehended A possible solution would be to add a label such as push both underneath the cancel symbol Also simultaneous keypresses would be physically easier if the buttons provided better feedback Problem 8 Several key labels were uninformative Many subjects did not understand the labels for cancel and enter One possibility for enter might be the graphic used on some computer keyboards 4 Problem 9 The keys are too small and too closely spaced This was a major problem Small keys on this device reduce the in vehicle real estate required lessening the degree to which the stalk mounted display unit blocks access 52 to other instrument panel controls and displays Compactness also reduces cost However the keys on the Ali Sc
63. l time time between keypresses and correctness of each destination designation This tool was developed because the Ali Scout interface is a closed product and connecting recording sensors to the Display Unit would have destroyed it For the simulation however keypress times and keys pressed were automatically recorded by the simulation software thus the data reduction software was required only for trials with the real Ali Scout unit The data reduction software developed VCRTimelt 1 0 controls a Panasonic AG DS550 video cassette recorder with single frame accuracy The interface see Figure 18 on the following page was written using SuperCard 2 5 for the Macintosh With VCRTimelt each time the user presses a key on the keyboard the character for that key along with the current VCR counter reading the time is added to a list field The user is free to play the tape at any speed and can choose any key to represent an event Normally the tape was played back at half speed with the analyst pressing an associated key at approximately the same time as a character appeared or the cursor moved on the Ali Scout display When subjects pressed buttons that produced no visual change to the display such as the shift key the analyst pressed the associated key close to the moment the subject released the button When the keypresses became frequent such as when the subject repeatedly pressed the same key playback was reduced to one quarter spe
64. location names street addresses and coordinates a focus of this experiment The review provided here is extremely detailed Those details concern subject samples tasks and test protocols all necessary to make comparisons of the relative ease of use of alternative interfaces One method to enter navigation information is to use a telephone keypad Figure 1 a topic addressed by Marics 1990 Keypads require a minimal of instrument panel space a premium in contemporary vehicles Marics examined behavior for entering names including q z apostrophe and hyphen characters not present on a keypad but present in names Twenty subjects were given a stack of index cards with 20 names on them and did what they thought was best to enter the names Table 1 summarizes the results Except for entering an apostrophe which subjects omitted no single method was preferred by more than half of the subjects This makes selecting a stereotype difficult For q and z the most commonly selected key was the asterisk selected by about 1 3 of the subjects ABC DEF 1 2 3 GHI JKL 4 5 PRSI TUV IWXY 7 8 9 OPER x Figure 1 Phone Keypad Table 1 Choices of Keys to Use in percent apostrophe 15 80 5 hyphen 1 35 35 30 In related work Detweiler 1990 examined alternative methods for entering text using a phone keypad Five methods were examined as described in Table 2 and Figure 2 Table 2 Descrip
65. looking at differences due to those factors See Appendix 1 There were differences for the individual entry tasks See Table 14 26 Table 14 Entry Lists for Each Stimulus Set Stimulus Set Name NICKS PLACE 12 1 GAS 7 2 HELENS KITCHEN 15 1 YAW GALLERY 12 1 GOODYEAR 8 0 FARMER JACK 12 1 TACO LOCO 10 1 FIRST OF AM 13 2 JACOBSONS 9 0 CHEVRON 8 1 LARK REST 10 1 UNICORN GRILL 14 1 KROGERS 7 0 QWIK STOP 11 2 TUFFY AUTO 11 1 LICHT PARK 11 1 NORDSTROM 9 0 DISCAFE 7 0 OAKLAND MALL 13 1 OLIVE GARDEN 14 2 The order of presentation of the three conditions simulated real display at dusk and real display at night was counterbalanced See Appendix G The subject s visual acuity was tested after the experimental blocks had been completed Then questions were asked regard ng key size key spacing display size display contrast and whether the system was log cal See Appendix J Subsequently subjects pressed their finger against a transparency on the face of a copier as if they were pressing a key A scale on the transparency aided in measuring the finger contact patch After completing the experiment subjects were thanked and paid 40 27 28 RESULTS Data reduction method An important contribution of this project was the development of software to obtain time and accuracy data from the videotapes of the trials for the real interface Specifically this software was designed to help obtain the tria
66. n a counterbalanced order Information requested included Points of Interest given the name find the coordinates in a list Address Ranges given an address find the coordinates in a list and Ali Scout Maps given two intersecting roads find the coordinates on a map Three different locations were examined for each information request The order of formats was counterbalanced across age groups See Appendix G After the practice and coordinate search tasks the subject was escorted to the driving simulator The road scene showed that the subject s car was parked on the side of the road After the driver adjusted the seat the experimenter verified that the subject could comfortably reach the touchscreen The interior lighting was adjusted for the test condition In the simulated dusk condition a small shielded light just above the center console and a second light in the passenger area were turned on The first block of trials consisted of five retrieval tasks followed by five entry tasks Appendix H lists the database used for retrieval tasks For each location in the database the minimum number of keystrokes needed to find the location was determined The only methods considered for this computation were pure scrolling alphanumeric and a simple version of the hybrid method the first letter followed by down arrows The 20 locations were split into 4 groups of 5 locations with the minimum number of keystrokes being equalized among
67. n the Eby Streff Wallace Kostyniuk Hopp and Underwood 1996 study of the Ali Scout interface more subjects rated the interface as easy to use than difficult to use but the responses were not overwhelmingly positive Almost half of the subjects disliked the keyboard to some degree Touchscreen simulations of user interfaces may offer reasonable performance predictions Movement times for alternative keyboard configurations can be predicted using equations based on Fitts Law For other information on design guidance see Green Levison Paelke and Serafin 1993 13 Research issues explored Thus while the literature suggests that GOMS predictions may be useful in evaluating the Ali Scout keyboard and that the basic Qwerty configuration is preferable the usability of the keyboard cannot be determined from the literature For this reason an experiment was conducted to determine if the interface was easy to use by drivers of all ages for destination entry and retrieval and to identify usability problems This experiment was carried out in parallel with the survey effort described in Eby Streff Wallace Kostyniuk Hopp and Underwood 1996 Specifically the following questions were addressed 1 2 What are typical entry and retrieval times and error rates for destinations How does the Ali Scout entry and retrieval times compare with those for other systems described in the literature Does performance change
68. n the destination s name longitude and latitude Tasks using the real interface were performed under both dusk and nighttime lighting conditions although the simulated interface was used only at dusk Some of the destination entry and retrieval times were unusually long As a result medians are more typical times than means although means were used for ANOVA comparisons Median retrieval times ranged from 0 4 to 12 0 seconds with an overall median of 6 2 Median entry times ranged from 39 5 to 67 6 seconds with an overall median of 51 5 An additional 30 to 60 seconds were required to look up coordinates in a manual Mean times for men were 34 percent longer for retrieval and 19 percent longer for entry than mean times for women both differences were statistically significant The ratio of mean times for older subjects to mean times for young subjects was 2 8 for retrieval and 2 2 for entry Performance also varied with context For retrieval the lighting condition was not significant but the simulated Ali Scout took about 75 percent longer than the real Ali Scout For entry times were 22 percent longer at night than at dusk 37 percent longer using the simulated interface Usability problems found involve labeling of keys the logic for shift key use and chanaing fields Key size and spacing and the lack of feedback were also concerns 17 Key Words 18 Distribution Statement ITS human factors ergonomics No restrictions
69. nd route mean 160 seconds and use a yellow page feature for a destination mean 90 seconds The error patterns were parallel to the task times Both sets of data were markedly affected by age for example the mean times for older subjects were almost double that of younger subjects Table 5 shows the full set of times for the sake of completeness Table 5 Task Completion Times from Dingus Hulse Krage Szczublewski and Berry 1991 Task Time s Enter an unfamiliar destination Retrieve stored destination Determine areas where congestion is present Use yellow pages feature to select a business Set Voice messaging option Summon emergency service One of the constraints of designing navigation interfaces is the limited space available on the instrument panel To address the issue of input device real estate requirements Sears Revis Swatski Crittenden and Shneiderman 1993 had 24 subjects type on a touchscreen Qwerty keyboard that registered input using a lift off strategy and presented a tone The four key sizes were 0 57 0 76 1 14 and 2 27 centimeters per side associated with keyboards 6 8 9 0 13 2 and 24 6 centimeters wide from the Q to P keys For each keyboard data was collected for a novice phase one practice string and three test strings and an experienced phase after 30 minutes of practice Mean entry rates ranged from 1 29 to 0 61 characters second for novices from smallest to largest keys and 0 57 to 0 37 for
70. o plan a route to your destination 86 00 find out how many miles you will drive on your route 25 44 access another function of the device 32 25 view your current location and your destination on the 8 44 same map find out how far it is to your next destination turn the voice down find out how far you have driven since the beginning of our trip replan your route to avoid Doolittle add yourself to the list of drivers modify your preferences by hiding the What to Do box change current driver to Clark add another leg to the plan Sales Calls 3450 96 44 Bayberry Rd Northbrook Delete the first left from the tnp plan Sales Calls Tailor the first leg by setting preference to min distance get help on the task you are doing now delete the trip plan named Steve s House delete the dr ver name Jon 26 89 select view from road as the format for your route 20 00 guidance get the next instruction 4 67 61 62 APPENDIX B DISPLAY UNIT LOCATION Top View face of CRT Front View y Eye brow 4 5 1 1 4 3 3 4 7 1 4 Found Button 3 4 n high x 1 1 4 wide Location of Actual Ali Scout unit 63 Note Since the Ali Scout display unit was mounted on a gooseneck and was moved aside periodically for other studies the location varied slightly from subject to subject plus or minus one inch Car approximate CR 12 2 Zr ME i 2 Simulated Display Uni
71. o small MEM large Is the text on the display the ric ur size too small too large 4 Does the screen have the right amount of contrast jnotenough just ght 2 much 9 Does the system behave in a logical manner no Simulated 00 5 APPENDIX K ANOVA TABLES ANOVA summary table for destination retrieval time Source df Sum of Squares Mean Square F Value P Value 1280602 1280602 7 769 0055 a 10027 110 5013 555 30414 0001 2 1078592 539296 3 272 0388 3d 15078253 502 008 3 049 0001 504 __ s83080785 164843 ANOVA summary table for destination entry time Source df Sum of Squares Mean Square F Value P Value 2 230400 068 115200034 108 882 5009 592 2504796 2367 3 220173228 73393408 6937 504 533244526 1058005 DO ANOVA summary of lighting condition effect and some interactions for destination retrieval The dependent variable is retrieval time dusk and night only Source df Sum of Squares Mean Square F Value P Value Subject Sex Age Group 32 2 503 Condition 1 11690 111690 1 243 Sex 01 316444 316444 3522 Condition Sex Condition Age Group 2 44 363 Residual 320 020 247 summary of lighting condition effect and some interactions for destination entry The dependent variable is entry time dusk and night only 87 Source df Sum of Squares Mean Square
72. oderately comfortable typing mean 3 9 42moderately comfortable Subjects were midway between very comfortable and moderately comfortable in using maps The subjects included both those who have served in previous UMTRI studies none of which concerned destination entry and new recruits New recruits were obtained through friends of the experimenters and other subjects Test materials and equipment Ali Scout interface This experiment incorporated a real Siemens Ali Scout Display Unit as well as a simulated version Figures 11 and 12 show frames grabbed from video recordings of device use Notice the similarity of the appearance of the two implementations These recordings were later used to identify subject actions and times for the real unit only Figure 11 The Real Display Unit Figure 12 The Simulated Display Unit 17 The Ali Scout interface consists of four elements an LCD guidance screen on the left of the unit face a text window for destination names coordinates entry information front panel selection keys and a fold out alphanumeric keypad Figures 13 and 14 actually taken from the simulation show the Display Unit closed and open These figures are full size on an 8 5 inch x 11 inch page Both the real and simulated Display Units were 7 1 4 inches x 2 1 4 inches 18 4 centimeters x 5 7 centimeters The found button was not part of the device but was added to assist in timing use performance The button was fabrica
73. ons sus trn tipi t uita bu uM AE 33 Did performance change with 35 How did performance time and errors vary as a function ol diver age and SON 37 What subject factors other than age and sex influenced performance pino 40 Were the times and errors the same for real and simulated Interfaces an 42 How accurate were subjects in looking up coordinates A CT DE 47 CONCLUSIONS AND 5 49 How can the analysis protocol be improved 49 What were typical entry and retrieval times and error rates f r DU ears 49 Did performance change with 49 How did time and errors vary as a function of driver age and sex 49 What subject factors other than age and sex influenced performance NAS expe iellE 50 How did time and errors vary as a function of ambient illumination 50 Were the times and errors the same for real and simulated interfaces seen een 51 What kinds of problems did drivers of all ages encounter and how can they be 51 How accurate were subjects in looking up coordinates Nte manuale 54
74. ons between the condition effect and the age effect The error bars represent 95 percent confidence intervals There were not enough errors committed during the retrieval task to analyze differences between the simulated Ali Scout and the real Ali Scout However the errors for the entry task were much greater in number for the simulated version than for either of the other two versions Figure 30 shows the condition effect identified by one way ANOVA as significant p 0 033 The 28 errors out of 180 trials committed on the simulated Ali Scout was at least twice the number committed in either of the lighting conditions for the real Ali Scout in the same number of trials 12 errors for the dusk condition and 14 for the night condition No interaction between condition and age group or sex was significant 43 1 2 ENTRY Errors committed out of 5 O Night Dusk Simulated Condition Figure 30 Effect of condition on errors committed The error bars represent 95 confidence intervals What kinds of interface design problems did drivers of all ages encounter and how can they be corrected The Ali Scout see Figure 31 has several usability problems and physical limitations some of which affected all subjects and some which affected primarily older subjects There were 11 problems worthy of note goes caer cee hee M Stee t ue Al B2 04 ES G7 H8 19 JO K Lu N 0 POR S T UU HX
75. out have about one quarter to one half the surface area of keys on a hand held calculator a device for which minimum key sizes are used for portability As indicated by the finger anthropometry data finger sizes of the subjects tested exceeded the width of the keys plus the surrounding space making it very difficult for subjects to precisely position their fingers and depress a single key Gloved operation important for Michigan winters would have been impossible Complaints concerning the keys appear in Eby et al 1996 and were voiced by subjects here The solution is straightforward increase the size and spacing The anthropometric data given in this report combined with the data in Hoffmann Tsang and Mu 1995 could be utilized to predict tradeoffs between entry performance times key size and key spacing However some adjustments of those data may be needed since the Hoffmann et al research assumes that subjects know where keys are located In the experiment conducted here some visual search for the key was required because of the nonstandard alphabetic sequencing Problem 10 Key feedback was inadequate Because there was not a distinct feeling when switch contact was made subjects waited for visual feedback from the real device and both visual and auditory feedback from the simulation to determine when switch closure occurred Because the feedback was not instantaneous subjects tended to press much harder and longer than necessary
76. ove average group only contained two Table 19 Median and mean times for above average and below average computer usage Computer Number of Subjects Above average 21 9 80 1 91 15 56 76 63 1 73 87 77 5 13 se 6 85 44 27 48 18 mputer W Below Avg Above Avg RETRIEVAL E interaction not 120 110 ENTRY interaction significant E 100 N 4 40 4 Mean Retrieval Time sec Mean Entry Time sec O T 2 lal aL Los Young Middle Older Young Middle Older Age Age Figure 27 Interaction between age and computer experience The error bars represent 95 percent confidence intervals 40 In an ANOVA of the number of errors committed in the destination entry task main effects of computer usage and age there were no significant differences due to computer usage F 1 32 0 245 p 0 6243 However the effect of age was significant F 2 32 7 594 p 0 002 consistent with results examining subject differences described earlier Without the age effect in the model the computer usage effect is barely significant p 0 047 mostly because the usage effect is confounded with the age effect How did time and errors vary as a function of ambient illumination Because the simulated interface shown on a CRT was self illuminated and legibility depended little on ambient illumination the simulated interface was not considered in the examina
77. pleting this study you can withdraw at any time You will be paid regardless e Have the subject read the consent form along with you In cars of the future you may have an in car navigation system which would tell you how to reach destinations To use it you ll need to tell the system where you want to go We are looking at how to enter and retrieve these destinations for a navigation system Responses from typical drivers such as you will help improve this system While sitting in a vehicle mockup you will respond to a simulated or a real navigation unit By pressing buttons destinations can be entered or retrieved A computer will record how long it takes to use the system With your permission we will videotape you We will not record your face at any time We will not release any identifying information so your responses will remain confidential e Have the subject fill out the consent and bio forms 71 PRACTICE The first part of this study is a short learning period to become acquainted with the navigation unit have a tape for you to watch that will provide a brief introduction Pay special attention to how the Ali Scout works and how the shift key is used e Watch the Ali Scout video e Hewind and eject the tape e Have the subject read the retrieval instructions e Insert the subject tape e Push Hecord e Turn off the big TV For practice please retrieve and enter the following sample destinations using the practice
78. provided with a reduced version of the manual a sheet with the command menu structure as previously shown and a practice sheet of instructions See Appendix F Subjects used a Display Unit to find a list of locations in a database that included five dummy and five real locations the Unused and Retrieved columns of Table 11 one at a time When all of the locations had been found the subjects were instructed to insert a list of five locations and their coordinates the Entered columns of Table 11 one at a time Table 11 Practice Databases Unused List Retrieved List Entered List dumm real Coordinates AMOCO BP SERVICE CADE GALLERY 0830845W 422908N BECKYS CAFE FIRESTONE HUNAN PALACE 0832531W 422805N ECHO PARK SCH MAYAS DELI MAIN THEATER 0830840W 422926N SIEMENS PLUS BANK 24 SHELL 0830532W 423534N STAR DELI SUBWAY VILLAGE MARKET 0830901W 423715N As can be seen from the listing of coordinates the length and content of all used coordinates are comparable across locations Therefore most discussion of entered locations will only focus on the names assigned to the locations The last portion of the activities conducted in the office was a coordinate search task for which name lengths were not comparable This required subjects to look up coordinates in the Ali Scout manual Location identifiers were listed on 3 x 5 inch 7 6 x 12 7 centimeter cards The experimenter handed the cards one at a time to the subject to maintai
79. r systems described in the literature Making comparisons with other studies in the literature is extremely difficult to do because the tasks and subject samples differ As was noted previously age differences have a major influence on performance with young old differences often differing by a factor of 1 5 Here when only keying times were examined the difference was a factor of 2 However interface differences tend to be much less For example in Paelke and Green 1993 the ratio of best to worst interfaces doublepress phone pad was 1 7 Fortunately some of the subject samples here are comparable Dingus Hulse Krage Szczublewski and Berry 1991 used three age groups young middle old while Paelke and Green 1993 had two groups of subjects young old Loring and Wiklund 1990b did not identify the ages of their subjects Interfaces examined varied widely in their functionality so functions provided on one system are not available on another Added functionality if not thoughtfully applied can make the completion of core tasks more difficult In several studies only a phrase is provided describing what subjects did so it is difficult to know if task objectives are comparable Finally there is some uncentainty about how tasks were timed While it is believed this generally occurred from the first button press planning prior to keying is an important part of destination entry and retrieval tasks However it is challengin
80. rall impression about 47 percent disliked the keyboard to some degree This is not a desirable outcome Several of the experiments described in this section used simulations of interfaces to predict performance with real interfaces To examine the validity of that approach Archer and Yuan 1995 presented four pairs of interfaces to 16 MBA students The interface examined various ways to enter text into a telephone system Two interfaces involved a 12 key phonepad supplemented with four additional keys alphabet number clear display send message and clear In the multipress implementation the character key is pressed to select that mode then the key with the character on it is pressed multiple times to select that character The 2 ABC key is pressed three times to select C This is similar but not identical to the repeat key method examined by Detweiler 1990 In the character pick interface pressing the alphabet key and then key for example 2 ABC cause the triple A B C to appear on a display To select C the 3 DEF key is pressed because it is the third key This is similar to the top row method examined by Detweiler 1990 Also examined were a Qwerty keyboard and a touchscreen simulation of one All simulations were implemented in Toolbook Three tasks were completed 1 ordering tickets using a data base with menus 2 entering credit card information to pay for the tickets and 3 entering an address to which the tickets were
81. rototypes in a laboratory Table 4 shows the mean times to complete tasks though the units are not given in the report The times are probably in minutes ES pz PSI EJ Ea E Select Backup Figure 3 Design Soft Keys and Dedicated Keys 12 dedicated keys including arrow cursor keys 12 key alphanumeric pad 5 soft keys and a menu hierarchy with many layers Get directions How to use Voice Next 2nd 3rd X 1 An HN Where am I Detour Pause Figure 4 Design B Touchscreen and Dedicated Keys 8 dedicated keys a touchscreen menu hierarchy with many layers and separate mini Qwerty keyboard Get directions Voice Format Howtouse More features Film OO O O Where am I Local Atlas map out Zoon in Detour Pause Backup Select route Select M Figure 5 Design C Dedicated Keys 18 dedicated keys including a cursor control knob a menu hierarchy with fewer layers and separate mini Qwerty keyboard Y Table 4 Task Times form Loring and Wiklund 1990a Task A soft keys amp B touchscreen amp C dedicated dedicated keys dedicated keys keys map publ places list of public places View traffic conditions near present 2 725 location San Francisco In structured interviews subjects said that design B appeared easiest to use initially proved easiest to use after one week and outranked the other two as the de
82. rp computer projection panel model QA 1650 O 3M overhead projector model 9550 Panasonic model GP KS152 lipstick camera with a 7 5mm lens D Power Macintosh9500 120 D Power Macintosh 7100 80AV JVC KM 1200 Color Special Effects Generator with a Panasonic AG 6200 Video Cassette Recorder D Lamp for dusk blocks Figure 17 Simulator Layout 2d Miscellaneous equipment In addition to the aforementioned equipment an Ikegami ITC 47 video camera a tripod and a Panasonic AG 1970 video cassette recorder were used to videotape the subjects from over their shoulders in the office during the practice and coordinate identification tasks illuminance was measured with a Minolta T 1 Illumination Meter The illuminance was measured at the location of the Display Unit Luminance was measured with a Spectra Pritchard Photometer models 1980A CD and OP The photometer was mounted on boxes and stabilized by ropes from above so that it was in the same position as the driver s eyes off to the side Some error may have been introduced because of the awkward posture required by the experimenter to aim and focus the photometer lean in through the driver s window and look through the viewfinder with the back of his head pressed against the B pillar of the car Test activities and their sequence After the initial greeting the subject received a general overview of the purpose of the study completed biographical and consent form
83. s contained in the Appendix D and E and completed a visual acuity test Following was a fixed set of activities learning how to use the device and practice with it the coordinate search tasks looking up addresses in books and the experimental destination retrieval and entry tasks The experiment ended with miscellaneous tasks to determine subjects impressions of the device See Table 10 for additional information See Appendix F for the complete instructions Table 10 Experiment Summary 8 Introduction subject is told purpose of experiment subject completes biographical and consent forms 2 Videotape subject watches instructional video on entering and retrieving destinations Test use of subject looks up 3 destinations in manual point of interest C name intersection of 2 roads street address Simulator subject is introduced to the touchscreen practice subject completes 5 entry then 5 retrieval tasks 3 times real use interface at dusk real at night simulated at dusk order was counterbalanced subject s eyesight checked subject completes questionnaire paid subject s finger anthropomet Post test is recorded subject 15 The learning and practice activities began with a five minute video tape on the Ali Scout system Wallace Eby and Gardner 1995 This tape included instruction on searching for coordinates and entering destinations Subsequently the subject was 24
84. s installed in their vehicles Ali Scout vehicles communicate with infrared roadside beacons which send travel times to the traffic control center and receive sequential routing instructions from the center If navigation products are to be produced they must be safe and easy to use The original program plan called for four human factors studies to examine safety and usability 1 destination entry and retrieval in the laboratory 2 route following on the road using the Ali Scout in an instrumented car 3 getting lost where drivers are taken off route to see how they and the navigation system recover and 4 a comparative evaluation of alternative navigation interfaces Study 4 was canceled first weakening study 2 as it was intended to provide baseline data for the Ali Scout Subsequently study 3 was canceled for lack of funding midway through study 2 During the original definition of the project the focus was evaluation of the Ali Scout interface with comparisons occurring in study 4 However as the project unfolded it became clear that a beacon based system with some of the limitations present in the Ali Scout interface was not likely to represent future products in the U S Further the cancellation of studies 3 and 4 meant that pilot comparison data had to be conducted in earlier studies so that the safety of the Ali Scout interface could be assessed Asa consequence of these changes emphasis was shifted towards a more general a
85. sign they preferred to have in their own cars They said that design C was the most difficult appearing initially and the hardest to use during the experiment Loring and Wiklund 1990b had 9 subjects varying in experience but of unknown age perform a variety of tasks with prototype 2 of the ADVANCE interface an illustration of the interface did not appear in their report Appendix A shows the adjusted mean times from that baseline evaluation The range of task times is quite large with one task taking over three minutes Coleman Loring and Wiklund 1991a b had 20 subjects enter alphabetic strings e g street names numeric strings e g long distance phone numbers and alphanumeric strings street addresses on 5 inch diagonal touchscreens Keyboard styles explored included Qwerty Qwerty matrix and alphabetic The two Qwerty keyboards were faster 0 55 seconds per keystroke than the alphabetic keyboard 0 73 seconds per keystroke Differences in errors were not statistically significant This suggests that where touchscreen resolution is available keys should be in a Qwerty format As part of the TravTek project Dingus Hulse Krage Szczublewski and Berry 1991 had subjects perform seven nondriving tasks with a navigation system interface Several of them involved entry or retrieval of destinations select an unfamiliar address mean 130 seconds select a stored destination mean 50 seconds store a destination a
86. simulated the number in the simulated condition was greater by a factor of 2 Even if the times and errors are different it may be possible to scale data from simulated interfaces to predict performance with real interfaces although interactions with age complicate scaling What kinds of problems did drivers of all ages encounter and how can they be corrected Problem 1 Zero and the letter O were confused These two characters look alike and were therefore mistaken for each other Solutions include adding a slash to the zero and placing a border around the numbers to group them together A more radical solution would be to completely redesign the keyboard so each key face is used for only one character Problem 2 Use of the shift was confusing While subjects were shown how to use the shift key some forgot In part this was because the toggling model for shift key use press shift then the key to be shifted did not fit the typewriter computer model hold the shift key down then press the key to be shifted One step in the right direction would be to have the display indicate whether the system was in shift mode via a light tone or LCD graphic Identifying the shift key was a problem The color differences between the white unshifted yellow shifted keys were not perceptible to some older drivers This aspect of the problem could be overcome in part by changing the font of one of the characters on each key or using a mor
87. ssessment of the desired qualities in navigation interfaces and protocols for assessing them Such shifts occurred without compromising the intent of the project as it was initially framed Driver navigation related tasks include 1 calibration and set up 2 telling the system where the driver wants to go destination designation and 3 following the guidance instructions The second and third tasks are more important The human factors work carried out in the FAST TRAC project is described in five reports Matters related to destination designation are covered in this report and a subsequent report on models the prediction of keystroke entry times Manes Green and Hunter 1996b Research relating to following route guidance is covered in three reports one concerning equipment used in the evaluation Katz Green and Fleming 1995 one concerning turn errors driving performance and subjective ratings Katz Fleming Hunter Green and Damouth 1996 and a third concerning driver eye glances Manes Green and Hunter 1996a in progress This particular report concerns the second task destination designation Relevant topics include how drivers determine the coordinates for a new destination how drivers enter coordinates into the navigation computer and how they retrieve previously entered destinations Times and errors for these tasks were collected In addition this report addresses a larger more fundamental scientific issue whet
88. t 4 3 4 10 RT ES RER SUR gt YE ard holder RARER RIES AOE 1 2 _ 73 4 Simulated display unit is roughly centered in display area Note card holder is 4 1 2 inh by 6 1 2 w Location of Simulated Display Unit shown on CRT 64 APPENDIX C LIGHT LEVELS FOR THE EXPERIMENTAL CONDITIONS lux Location Used Color Background white 15 10 1 00 J white 7 17 2 47 Simulated Dusk 31 8 yellow 7 40 2 47 white 9 83 1 83 i T black 1 14 356 white 230 0419 i J white 1 03 0 24 Dusk 28 2 O yellow 1 19 0 24 white 2 43 0 25 3 03 dorup white 086 0 06 En dn white 0 44 0 09 Night 14 7 vellow 1 30 0 09 white 1 30 0 15 display T 1 54 2 62 note the T from the display is at the starT 65 66 APPENDIX D BIOGRAPHICAL FORM University of Michigan Transportation Research Institute Human Factors Division Subject Biographical Form Dae Male Female circle one Age Occupation Education circle highest level completed some high school high school degree some trade tech school trade tech school degree some college college degree some graduate school graduate school degree Other If retired or student note it and your former occupation or major What kind of car do you drive the most Year Make Model Annual mileage Have you ever driven a vehicle wit
89. t difference between women s and men s retrieval and entry times is undoubtedly the result of many factors but one which may have some influence is finger size Figure 32 shows the results of the anthropometry analysis in which each subject s finger contact width was measured The width of an Ali Scout button is 6 0 millimeters and the space between adjacent buttons is 3 0 millimeters This suggests that 12 0 millimeters is the maximum width of a finger if no contact with other buttons is to occur Of the subjects in this study 11 of the 18 women and 16 of the 18 men had finger contact patches wider than 12 0 millimeters In the judgment of the experimenters subjects in this sample of adults did not seem to have unusually fat fingers Women Men 6 A 6 Mean 125mm Mean 13 8 mm 57 A Stdev 1 4 mm St dev 1 7 mm Ber 9 2 5 O 5 2 1 E 1 10 12 14 16 18 20 10 12 14 16 18 20 Finger width mm Finger width mm Figure 32 Distribution of finger contact widths for the 36 subjects split by gender Normal curves with the same mean and standard deviation are overlaid Problem 10 The lack of feedback from the key mechanisms made keying difficult In particular since there was no distinct bottoming point subjects did not always provide the proper amount of pressure There were also many instances where subjects pressed and released a key and then wait
90. ted out of wood and attached to the zero key of a Kensington NoteBook KeyPad mounted just behind the Ali Scout unit Figure 13 The Simulated Display Unit with the Door Up 18 NE RN RD ND Al B2 D4 ES F6 67 19 JO K Z A Luh N 0 POR 5 T UU HX YZ DO J3 A b NH A LE LEI o Figure 14 The Simulated Display Unit with the Door Down The simulated Display Unit was created in SuperCard 1 7 1 The program recorded the times for button presses and opening and closing the door The simulated image of a Display Unit was presented on a ELO Touch Systems Intellitouch monitor model E284A 1345 mounted on an Apple Macintosh model M1212 13 inch 33 centimeter CRT positioned in the center console of the dashboard The frame of the display is visible in Figure 15 The size and appearance of all elements of the simulated display were identical to the real interface except that there was no tactile feedback when a key was pressed something that created major problems for subjects in pilot tests As a consequence in the simulation a tone was presented each time a key was pressed This seems to have solved the problem The real Display Unit was mounted on a flexible stalk that was positioned by the experimenter placing the display face 3 1 2 inches 8 9 centimeters in front of the touchscreen when in use A holder for the address cards was mounted just below the display Dimensions for the location of
91. than those for retrieval for a couple reasons First because retrieval times are shorter the lengths of time spent before the first keypress and after the last which are missing from the times recorded in Table 20 are a greater proportion of total retrieval time than of total entry time Second retrieval times particularly if the subject scrolls through the whole list to find the desired location are dependent on the number and sequence of destinations stored in the memory of the Ali Scout whereas entry times are not Table 16 gives the total number of errors for each destination for the 36 subjects To determine error rates divide the number of errors by 36 Because entry involves typing data into the Ali Scout and retneval only involves finding data that are already there entry errors were much more common than retrieval errors As noted earlier these errors are only of the type that were never corrected by the subjects Table 21 does not for example include information on missed keypresses that were immediately corrected 34 Table 16 Numbers of errors for both the retrieval and entry cards Retrieval Task Errors Entry Task Errors Place Name Errors Place Name Errors SAKURA BANK 0 NICKS PLACE 6 BIR ICE ARENA 1 GAS 4 MONTERREY REST 1 HELENS KITCHEN 6 MOBIL 0 YAW GALLERY 4 BIG BOY 0 2 SEAFOOD BAY 0 2 PRINT GALLERY 0 1 MAJESTIC CAFE 0 4 VANDENBURG SCH 1 4 LIBRARY 5 6 BILL KNAPPS 0 0 PRIMOS P
92. tion of Methods Used by Detweiler 1990 Method Description to enter K Hepeat key press 5 twice once to indicate which group of three letters a second time because K is the second of 3 letters JKL Same row press 5 twice once to select the JKL key a second time because 5 is the second key on the same row GHI JKL MNO press 5 to select the JKL key then 2 on the top row because K is the second of 3 letters JKL Modified modal press OPER to select the second position it is the second key in on the bottom row then 5 because JKL 15 on that key This method Note The general approach is the first keystroke selects the key subsequent keystrokes indicate the specific character on the key resembles the same row and top tow methods except that it uses the bottom row and the keys are reversed press 5 to select the JKL key then the first special character to select K to select J press only 5 to select L press 5 to select the JKL key then the second special character to select L Repeat Key Method JKL JKL TUV TUV DEF GHI GHI 5115 19 1818 18031414 Same Row Method JKL JKL TUV WXY DEF GHI GHI MNO 5 5 9 7 8 9 3 1 4 4 6 6 Top Row Method JKL ABC Wx DEF DEF GHI ABC MNA DEF 5 2 19 11118 114 1 21 6 3 Modal Position Method OPER JKL WX TUV DEF G
93. tion of the effects of ambient illumination Thus this section compares only the dusk and night conditions for the real Ali Scout interface For destination retrieval the mean times were 8 89 seconds for the night condition and 7 77 seconds for the dusk condition This difference is not significant pz0 2657 nor is the interaction of lighting with age p 0 8863 or sex p 0 7814 significant These results are revealed by the ANOVA summary in Appendix K For tables of the median values see Table 20 in the next section However for destination entry the same ANOVA model reveals a significant difference between dusk and night p 0 0001 as well as a significant interaction between the lighting effect and the age effect p 0 0001 See Appendix K The interaction plot in Figure 28 reveals that the observed significant difference between the night and dusk conditions for the real interface night meanz65 7 sec dusk mean 54 0 sec is mostly attributable to the older subjects In the night condition many of the older and some of the middle aged subjects were unable to see the top row of the keypad because it is unlit This forced them to count over to the key they were looking for starting with the first or last key in the row When attempting to enter a zero subjects were especially likely instead to attempt to enter the letter O since that is the only round symbol they could see or a K which was the last key of the top row 41 12
94. to be sent Subjects then rated the pair of interfaces they used on 10 characteristics Table 7 shows the entry times and errors The touchscreen simulation took 14 percent longer and led to 50 percent more errors than the real interface However the actual number of characters entered was only 3 percent greater The time differences were not statistically significant leading the authors to claim that the use of simulations is an appropriate substitute for real interfaces in usability tests Table 7 Entry Times and Errors Reported by Archer and Yuan Not only are evaluations of real interfaces important but so too are evaluations that develop a basis for predicting user performance As an example Hoffman Tsang and Mu 1995 described two experiments to predict movement times between keys as a function of key size and spacing In the first experiment 10 young men served as subjects The square keys were 5 10 and 15 millimeters on each side separated by either 1 5 10 15 or 20 millimeter gaps Reciprocal movements were between a starting key and a second key 1 to 5 keys to the right The performance measure was the number of key taps in 10 seconds The best fitting equation for movement time milliseconds accounting for 95 percent of the variance was 12 MT 187 6 68 ID 1 18 2 13 3 Nkey ID Index of Difficulty 2 movement distance target width Nkey number of keys to the second key target width key width
95. unit e Provide the practice sheet and paired manual Retrieve List Enter List BP Service Cade Gallery 0830845W 422908N Firestone Hunan Palace 0832531W 422805N Mayas Deli Main Theater 0830840W 422926N Plus Bank 24 Shell 0830532W 423534N Subway Village Market 0830901W 423715N Echo Park Sch Siemens Star Deli 72 COORDINATE FINDING To understand how people find coordinates for this navigation system we are going to have you use the manual to find some navigation coordinates We will be videotaping you from an angle that will not show your face Please do not write in the manual e Run the coordinate finding series Coordinate List List Format Ali Scout Maps WER On Map 1 Giddings amp Walton 0831540W 4240404 On 2 Auburn amp Ryan 0830420W 423820N On Map 3 Franklin amp Maple 0831810W Address Ranges 450 12 Mile Rd W 2300 Coolidge Hwy 4429 Sashabaw Rd Points of Interest Troy Union School Glen Oaks Golf Course Nicholas George Theater near middle 1 2way to edge 423240N bottom edge 0830635W 423017N 0831101W 422939N 0832212W 424211N middle of range start end 0830722W 423625N 0832116W 423050N 0831425W 422803N e Eject the tape Turn off all audio video equipment 73 Coordinate Pattern O E UO amp kon o o O Demographics 1 Group
96. use they were generated by a CRT were back illuminated and did not require ambient illumination The worst case was an alphabetic character J on the door contrast ratio 2 9 1 There were numerous situations in which contrast ratios of 10 1 were achieved Appendix C contains the illuminance and luminance values for the various experimental conditions Driving simulator Interaction with the subjects occurred at two locations in UMTRI Training and the collection of subject information was conducted in an office The data collection portion of the experiment was conducted in the Driver Interface Research Simulator 22 The automobile simulator consisted of an automobile cab a retroreflective wall and a variety of computer and video components Figure 17 is a simplified drawing of the equipment used in this particular study It is important to note that subjects never drove the simulator It simply provided context During experimental trials a road scene was presented simulating sitting on the right shoulder of a two lane road M 12 Q 1985 Chrysler Laser mockup with simulated hood 8 X10 projection screen with 3M hi white 20 encapsulated reflective sheeting 6 Ali Scout Display Unit on a flexible stalk with a custom extemal power supply 3 spoke steering wheel Q Sharp color LCD projection system model XG E850U 4 X13 plexiglas screen ELO Touch Systems Intellitouch monitor model E284A 1345 Sha
97. vels Appendix K shows the same two models for the dusk night comparison but with the additional data for the simulated Ali Scout included The same qualitative results are seen for both the retrieval times and the entry times The condition effect is significant in each p 0 0001 and the interaction between condition and age group is significant in each p 0 0001 In Table 20 the median and mean times are given for each of the three conditions Although the differences between simulated times and other times are significant for both entry and retrieval the differences are larger for the retrieval times In the interaction plots shown in Figure 29 the main qualitative difference between the retrieval and entry tasks occurs with the older subjects In the retrieval case the unique condition is the simulated condition which is much longer in the entry case the unique condition is the dusk condition which is much shorter 42 Table 20 Median and mean times for each condition Destination Retrieval s Median Ratio to Sim Simulated 8 42 1 00 Night 6 03 072 Dusk 4 79 0 57 Destination Entry s Median Ratio Simulated 58 65 1 00 Night 50 32 0 86 Dusk 44 52 0 76 RETRIEVAL ENTRY 12 2 30 S O E25 o 100 20 E 80 e 15 E 60 10 c 5 2 40 gt 20 Young Middle Older Young Middle Older Age Age Figure 29 Interacti
98. with practice How does performance time and errors vary as a function of driver age and sex What subject factors other than age and sex influence performance in this experiment How do time and errors vary as a function of ambient illumination Are the times and errors the same for real and simulated interfaces What kinds of problems do drivers of all ages encounter and how can they be corrected How close are GOMS predictions of times to values from real and simulated interfaces 10 How accurate are subjects in looking up coordinates in the manual The rationale for these questions is shown in Table 14 Table 8 Rationale for the Issues Examined Rationale Deeper Issue Typical times and errors To make comparisons both with other systems and to gauge improvements baseline data are needed Comparison with other How should a navigation system be designed to be safe systems and easy to use Effects of age sex and Will any group have particular problems This influences other subject factors who should be subjects in future usability tests Ambient illumination Are there lighting problems Does lighting in tests matter Should multiple lighting conditions be examined in future tests Real vs simulated Can simulated prototypes popular in preliminary tests interfaces predict real system usability This could reduce the cost of usability tests needed during development since only a simulation is need
99. y Retrieval Pattern Demographics Age Group 1 1 DO Destination Retrieval Entry Blocks 1 Cards 2 Cards 3 Cards Light B Light C Dark A Sim A Light B Dark B Sim C Dark B Light A Light C Sim B S m B Light C Dark Sim A Dark A Light Sim C Dark Light Sim Light B Dark Sim C Light Dark A Sim B CONCLUSION The data entry tasks are completed would now like to check your eyesight e Make sure eyesight condition is the same as when testing was done e Test the subject s eyesight We will now go back to the office to fill out some final forms e Walk the subject back to the office Please fill out this question sheet If you have additional comments please use the space provided e Have the subject fill out the question sheet Please fill out this form so that can pay you e Have the subject fill out the payment form e Walk the subject towards the stairs e Pay the subject Before you go would like to measure your finger size so that can evaluate the size and placement of the buttons on the navigation unit e Walk the subject to the copy room Make a copy of the subject s finger using the cardboard with plastic window AFTER SUBJECT LEAVES Sim Lab e Unplug the microphone e Tum off the two computers e Remove the duct tape from the door Tavern e Clear
100. younger subjects and women vs men e Entry and retrieval times were significantly longer for night condition vs dusk condition trials and simulated interface vs real interface trials The Ali Scout keys are too small too close together and do not provide enough positive feedback when pressed Some labels should be improved The shifting and spacing functions are confusing The results raise major concerns regarding the usability of any on road navigation system that relies upon longitude and latitude for destination ID IV PREFACE This report is one of a series supported by the Road Commission of Oakland County Michigan and the Federal Highway Administration as part of the FAST TRAC Faster and Safer Travel through Traffic Routing and Advanced Controls Project See Underwood 1994 Eby Streff Wallace Kostyniuk Hopp and Underwood 1996 Taylor and Wu 1995 Kostyniuk and Eby 1996 for related research This operational field test combines the SCATS Sydney Coordinated Automatic Traffic Control System equipment and software the Autoscope video detection system and the Ali Scout Autofahrer Leit und Information System Scout dynamic route guidance system The goals of this effort are to improve traffic flow and reduce traffic accidents in Oakland County and the surrounding area Ali Scout is a second generation product developed by Siemens which provides real time turn by turn guidance to drivers who have unit

0 - Deep Blue - University of Michigan

Contents

Download Pdf Manuals

Related Search

Related Contents