raskin_chp3_4

raskin_chp3_4 - 70 MEANINGS. MODES. MONOTONY. AND MYTHS...

Info iconThis preview shows pages 1–14. Sign up to view the full content.

View Full Document Right Arrow Icon
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 2
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 4
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 6
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 8
Background image of page 9

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 10
Background image of page 11

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 12
Background image of page 13

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 14
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: 70 MEANINGS. MODES. MONOTONY. AND MYTHS system changes automatically, even if the change is as small as, say, a reordered set of items on a menu, your expectations are upset and your habituation is frustrated. (Microsoft features adaptive menus in its Windows 2000 operating system?) On the other hand, there is no theory that tells us that the same fixed interface cannot work well over the full span of a per~ son’s experience with it, from novice to old timer. It seems best not to have to shift paradigms during your use of a product, and no elaborate analysis is needed to reveal the advantage in having to learn only one interface to a task. It is easy to fall into the trap of designing different‘interfaces for diff ferent classes of users, because by doing so, you can make sweeping assump— tions that simplify the design process. Few such assumptions are likely to be true of every user in any reasonably large class of users that you specify. The antidote is to view an interface not from the perspective of a class of users but rather through the eyes of an individual. Every person who uses software over a long period goes through a relatively brief period of learning for each feature or command and a far longer period of routine (and, we hope, auto— matic) use. We do need to design systems that are easy to learn and under— stand, but it is more important that we make sure that these systems can be efficiently used in the long run. The exceptions are applications that will be used only briefly, so that every user is a novice, and habituation is not an issue. One example of such an interface is that for a computeridriven kiosk at an exhibition. The learning phase of working with a feature involves your conscious attention. Therefore, simplicity, clarity of function, and visibility are of great importance. The expert phase is predominantly characterized by uncon— scious use of the feature; such use is enhanced by such qualities as aptness to the task, modelessness, and monotony. These sets of requirements are not in conflict; therefore, a well—designed and humane interface does not have to be split into beginner and expert subsystems. This is not to say that an interface must not be split on these lines. However, if you find yourself designing an interface and are tempted to pro— vide “expert” shortcuts, consider whether you should instead redesign the existing method so that it satisfies the needs of all users with one mechanism. 9. Windows 2000 was a new product as this section was being written, and l was able to inter- view only a few users. A typical remark was, “Adaptive menus seemed like a cool idea, but the first tilne a menu changed on me, I found it upsetting. I don’t like the idea any more,” FOUR Quantification The harmony of the world is made manfest in Form anti1 Num- ber, and the heart and soul ana1 all the poetry of Natural Philos- ophy are embodied in the concept of mathematical beauty. #D’Arcy [M’ntworth Wronrpson, On Growth and Form (1917) A number of methods for analyzing interface details quantitatively are available. However, explicit directions on how to use them are rare. This chapter gives an easyetouuse treatment and fully worked—out examples of Card, Moran, and Newell’s keystroke—level GOMS model, Raskin’s measures of efficiency, Hick’s Law, and Fitts’ law. 4-1 Quantitative Analyses of Interfaces He fingered andfingered the computer—it simply amazed Melrose that the machine supposed to take the pain out qfall sorts qugglingjobs took more time to perform a simple one than it would have taken Bub to do by hand ten times over. #Martha Crimes, The Stargazey (a detective novel) Many qualitative methods and heuristics are useful for analyzing and understanding interface design. These methods form the majority of the 71 72 QUANTIFICATION content of most books on the subject, including those cited in the references for Shneiderman, Norman, and Mayhew. For example, what an experifi enced interface designer can learn from passively observing a test of a new interface with a few subjects can be as valuable as what she can learn from any quantitative analysis. My concentration on quantitative methods is not meant to denigrate the importance of qualitative techniques but rather to help even the balance by emphasizing the numerical and empirically testable methods that are not yet widely used. Quantitative methods can often reduce argument to calculation; 3 further, and most important, benefit is that understanding why the quantitative methods work guides us to understanding impor- tant aspects of how humans interact with machines. ‘ One of the best quantitative analyses of interface design is the classic model of goals, objects, methods, and selection rules (GOMS), which first gained attention in the 19803 (Card, Moran, and Newell 1983). GOMS modeling allows you to predict how long an experienced worker will take to perform a particular operation when using a given interface design. After discussing the GOMS model, I present quantitative methods for determin— ing interface efficiency, cursor movement speed, and the time cost of deci— sion making. 4-2 GOMS Keystroke—Level Model The aim of exact science is to reduce the problems of nature to the determina- tion qf quantities by operations with numbers. #james Clerk Maxwell, On Faraday’s Lines of Force (1856) I will introduce only the simplest—yet nonetheless vfluableJaspect of the GOMS method: the keystroke—level model. We designers who know GOMS rarely use a detailed and formal analysis of an interface design, but that is due, in part, to our having absorbed the fundamentals of GOMS and of other quantitative methods such that our designs inherently in— corporate GOMS teachings. We do bring formal analysis into play when choosing between two approaches to interface design in which small dif— ferences in speed can have significant economic or psychological effects. We can sometimes benefit from the impressive accuracy of the more complete GOMS models, such as critical—path method GOMS (CPMiGOMS) or a version called natural GOMS language (NGOMSL), which takes into account nonexpert behavior, such as learning times. We can, for example, predict how long it will take a user to execute a particular set of interface actions to Within GOMS KEYSTROKE-LEVEL MODEL 73 an absolute error of less than 5 percent. In these advanced models, almost all predictions fall within 1 standard deviation of the measured times (Gray, John, and Atwood 1993, p. 278). In a field in which religious wars are waged over interface designs and in which gurus often have widely varying opin— ions, it is advantageous to have in your armamentarium quantitative, experi— mentally validated, and theoretically sound techniques. For a good overview and bibliography of the various GOMS models, including her own CPM— GOMS model, see John 1995. 4-2—1 Interface Timings Numerical precision is the very soul ofsrience. —D’/-lrcy VVentworth Thompson, On Growth and Form (1917) When they developed the GOMS model, its inventors observed that the time it takes the user—computer system to perform a task is the sum of the times it takes for the system to perform the serial elementary gestures that the task comprises. Although different users might have widely varying times, the researchers found that for many comparative analyses of tasks involving use of a keyboard and a graphical input device, you. could use a set of typical times rather than measuring the times of individuals. By means of carefiil laboratory experiments, they developed a set of timings for different gestures. In giving the timings, I follow the original nomenclature, in which each of the times is designated by a one—letter mnemonic (Card, Moran, and Newell 1983): K = 0.2 sec Keying: The time it takes to tap a key on the key— board P : 1.1 sec Pointing: The time it takes a user to point to a posi— tion on a display H = 0.4 sec Homing: The time it takes a user’s hand to move from the keyboard to the GID or from the GID to the keyboard M = 1.35 sec Mentally preparing: The time it takes a user to pre— pare mentally for the next step R Responding: The time a user must wait for a com— puter to respond to input in practice, these numbers vary widely; K can be 0.08 sec for a 135— me highly skilled typist, 0.2 sec for a more typical SS—me skilled typist, 74 QUANTIFICATION 0.28 sec for a 40—wpm average unskilled typist, or 1.2 sec for a beginning typist. Typing speed is not independent of what is being typed: It takes most people about 0.5 sec to type a random letter, given a set of randomly chosen letters to type. Typing messy codesfifor example, e—mail addresses—takes most people about 0.75 sec per character. The value K includes time it takes the user to make corrections that he has caught immediately. Shift is counted as a separate keystroke. The wide variability of each measure explains why we cannot use this simplified model to obtain absolute timings with any degree of cer— tainty; by using the typical values, however, we usually obtain the correct ranking of the performance times of two interface designs. If you are evalu- ating complex interfaces that include overlapping time dependencies or if you must generate accurate absolute times, you should use the more com— plete models, which are not discussed in this book, such as CPM-GOMS. Double Dysclicksia The interface technique called double clicking, that is, tapping the GID button twice within a small time window and without any significant cursor movement between the taps, as an interface tech— nique suffers from problems. You cannot always predict what objects on the display will or will not respond to a double click, and it is not always clear what will happen if there is a response. There is no indi— cation on displayable items that double clicking is supposed to pro— duce a response: The fiinctionality is invisible. The way that double clicking is used in many current interfaces, the user must remember not only which items are double clickable but also how different classes of interface features respond to this action. The first two burdens on the user could be at least partially alleviated by new screen conventions. The act of double clicking is, however, itself problematical. Double clicking requires operating a mouse but" ton twice at the same location or at two locations in very close and, in most cases, within a short time, typically 500 msec. If the user clicks too slowly, the machine responds to two single clicks rather than to one double click. If the user jiggles the mouse excessively between clicks, the same error occurs. If the user taps the GID button twice in too short a time period, as when trying to select text within a word while working within certain word processors, the machine considers the two taps as a double click and Selects the whole word GOMS KEYSTROKE-LEVEL MODE 75 A problem arises when the user is trying to select a graphical item that can be repositioned with the CID. Because the GID is likely to move when the user is pressing the GID buttons quickly, graphical applications, instead of reading a double click, may read a drag—and— drop and change the item’s position. Similarly, to change the text in a text box, the user may find it necessary to reposition the accidentally moved box and to make the text edit originally intended. some of us are unaffected by dysclicksia: These lucky people never miss with the mouse; they single and double click with insouciance and panache, do not suffer from side effects of clicking, always remember what will and what will not respond to double clicking and can shoot a flying bird with a .357—caliber revolver while driv: mg along a twisty mountain road. But we can’t assume that all users are so lucky. We must design for the dysclicksic user and remain aware of the problems inherent in using double clicks in an interface.1 The duration of the machine response time, R, can have an unex— pected effect on user actions. If a user operates a control and nothing appears on the display for more than approximately 250 msec, she is likely to become uneasy, to try again, or to begin to wonder whether the system is failing. We cannot build products that can complete any operation within human reaction time, but our interfaces can aIWays, within that time give feedback that the input has been received and recognized. Otherwise, user actions—often flailing at the keyboard, trying to get a response—during a delay can start the system off on unintended activities, causing fiirther delay or damaging the user’s content. For example, if you try to download a file while aceessmg America Online from a browser, such as Netscape’s there is often a long delay. No feedback lets you know that progress is being made' a small Static'message far from the locus of attention says only that the computer is Exaltlng fieply.‘ After a few seconds, inexperienced users start clicking at but— 5 on t e display which stops the download—again without feedback. It is important that interfaces provide feedback if delays are unavoid— b . . . a 16, display a progress bar (Figure 4.1) that accurately reflects the time remai ' ~ ' ' I fling. If you cannot predict how much time an operation will take say so. Do not he to or niisinfortn users. , -‘__—__ 1 Th e ‘ . , . . term dysditi'esia, a disease for which the only permanent cure is good design was coined by Pam Martin (personal communication 1997) 76 CLUANTIFICATION Fi are 4.1. A progress bar. It is important that it represent timeI 1:1; willy A textual statement of time remainmg, 1f accurate, £5 a s humane feature when delays are unavoidable. 4—2-2 GOMS Calculations We begin the calculation of the time it takes to perfori: a grief):sz such as “move your hand From the graphical input dev1ce golt efeyistulres and type a letter,” by listing the operations from the COM _ 15!: IC: gesmres (see Section 4—2-1) used in this method, 1n this case HK,List1nSg t edg15 The (K, P, and H) is the easy part of creating an instance of GOM 1mo1 cc;sz more difficult part of developing an instance 'of a keystrolpe— eve num— model is figuring out at what points the userpvvfll stop. to per lprrlr: an puiss— scious mental operation: the mental preparation (M) t1mes.T 3e 351:65_f0r following the methods of Card, Moran, and Newell 198 , pd in Table deciding where mental operations occur 11’] a method are presente 4 1 In Section 4—23, we look at how these rules are applied in practice. . a I In these rules, a string is a sequence of characters. A delimiter is character that marks the beginning or the end of a meanmgful string of $2: such as a natural—language word or a telephone number. For exangple, spar at are the delimiters For most words; a period is the most common I e 121m eon the end of a sentence; parentheses delimit parenthetical remarks, an some}; The operators are K, P, and H. When a command needs infortmatgogfi and as when you use the command that sets the time for an alarm ofignt for that have to supply the time, the information you supply is an argum command. 4—2-3 GOMS Calculation Examples 3 e y § W I k (II 56‘ (IE tQS 3 All Inter CE d Sign Lsuall beglnc 1t 3 Has a I k 3(1 0- e U 5 e t t e tOl t C 5 a :11 3. I 3 em n h task and tile llieall V 1 ble 1’16 1: b 3CC ill-11311.11 A S I I] ll 3 e (l 160M611 (l Hulls-ted as a [BqUIICIHBIlt O SpCClllca 1 1m ntl gab utlon a I l 6.01]. 111 thls examplt, the user is personified 9.5 a labO‘latOry aSSIStant. GOMS KEYSTROKE-I.EVEL MODE TABLE 4.1. HEURISTICS FOR PLACING MENTAL OPERATORS Rule 0 Initial insertion of candidate Ms Insert Ms in front of all Ks (keystrokes). Place Ms in fiont of all Ps (acts of pointing with the GID) that select commands, but do not place Ads in Front of any P5 that point to arguments of those commands. Rule 1 Deletion of anticipated Ms If an operator following an M is fully anticipated in an operator just previ~ ous to that M, then delete that M. For example, if you move the GID with the intent of tapping the GID button when you reach the target of your GID move, then you delete. by this rule, the Myou inserted as a conse— quence of rule 0. In this case, PMKbecomes PK. Rule 2 Deletion of Ms within cognitive units If a string ofMKs belongs to a cognitive unit, then delete all the Ms but the first. A cognitive unit is a contiguous sequence of typed characters that form a command name or that is required as an argument to a command. For example, Y, move, Helen (JfTray, or 4564.23 can be examples of cognie rive units. Rule 3 Deletion of Ms before consecutive terminators If a Kis a redundant delimiter at the end of a cognitive unit, such as the delimiter of a command immediately following the delimiter of its argu— ment, then delete the Min front of it. Rule 4 Deletion of Ms that are terminators of commands If a K is a delimiter that follows a constant strinpfor example, a com— mand name or any typed entity that is the same every time that you use itithen delete the M in front of it. (Adding the delimiter will have become habitual, and thus the delimiter will have become part of the string and not require a separate M.) But if the K is a delimiter for an argument string 01' any string that can vary, then keep the M in firont of it. Rule 5 Deletion of overlapped Ms Do not count any portion of an M that overlaps an R—a delay, with the user waiting for a response From the computer. Requirement Hal works at a computer, typing reports; he is occasionally inter— rupted by one or another of the researchers in the room, and is asked to convert a temperature reading from degrees Fahrenheit (F) or Celsius (C) to degrees C or F, respectively. For example, Hal might be asked, “Please convert 302.25 degrees from Fahrenheit to 77' 7s OCUANT'F‘C’G‘TION Celsius" Hal must use the keyboard or GID to enter the tempera— ture provided; voice or other input means are not available. Conver— sions from C to F and from F to C are approximately equally likely to be required. About 25 percent of the temperatures called out are negativo, although the digits are unpredictable and equally dise tributed, and only 10 percent of the temperatures have integer val— ues, such as 37 degrees. The numerical result must appear on the display; no other output means are available. Hal reads to the researcher the converted value from the screen. The input and the output must allow for at least ten digits on each side of the decimal point. l In designing an interface for a system that allows Hal to do his job, your goal is to minimize the time it takes Hal to do the conver— sion. Speed and accuracy must be maximized; screen real estate is not limited. The window, or area of the display in which the tempera— ture conversion takes place, is already active and waiting for Hal’s input via GID or keyboard. The way Hal interacts with the interface to return to his typing on the computer is not your concern; your job is finished as soon as the result is displayed. In estimating the time it takes Hal to use the interface, assume an average of four typed characters in an entered temperature, includ— ing any decimal point and sign. Also assume—unrealistically, but for simplicity’s sake—that Hal’s typing is perfect; error detection and notification are not needed. Now, I would like you to stop reading so that you can design an interface for this simple example. It will not take long to write down your proposed solution, along with sketches of the display that Hal will see; do not just think about this problem but rather write about it as well. (You will be tempted to read on without honoring my request. Please reconsider. The next few sections will make much more interesting reading if you have already tried to solve the problem yourself.) After designing your interface, read the two GOMS analyses that follow. Then you will be ready to analyze your own interface. 4-2-3—1 Hal’s Interface: Solution 1, Dialog Box The instructions in Figure 4.2 are reasonably clear; from them we can write down the method that Hal must use in terms of the gestures of the GOMS model. The GOMS representation is shown growing incrementally as each new gesture is added to the method. GOMS KEYSTROKE-LEVEL MODE 79 Temperaturgoter Choose which conversion is desired, then type the temperature and press Enter. © Convert F to C 0 Convert C to F ::> Figure 4.2. A dialog box solution with radio buttons. - Move hand to the graphical input device: H ° Point to the desired radio button: H P - Click on the radio button: H P K ‘ Half of the time, the interface will already have the correct conver— sion chosen, and Hal will not need to click on the radio button. We consider first the case in which it is not the one already chosen. - Move hands back to the keyboard: H PK H ' Type the four characters: H P K H K K K K ' Tap Enter: H P K H K K K K K t The keystroke for the tap of the Enter key completes the method portion of the analysis. Using rule 0, we add Ms in front of all of the Ks and PS except those Ps that point to arguments, of which there are none in this example: HMPMKHMKMKMKMKMK Rule 1 tells us to change P M K to P K and to eliminate any other fiilly antimpated Ms, of which there are none in this example. Rule 2 eliminates 8O QUANTI FICATION Ms in the middle of strings. such as in the string that represents the tempera— ture. Applying these two rules leaves HMPKHMKKKKMK The M before the final K is required by rule 4. Rules 3 and 5 do not apply in this example. The next step is to add the times represented by the letters. (Recall that K: 0.2, P =1.1, H: 0.4, and M: 1.35): H+M+P+K+H+M+K+K+K+K+M+K= 0.4 +1.35 + 1.1 + 0.2 + 0.4 +1.35 + 4*(0 2).|£1135 + 0.2 = 715 seconds In the case in which the correct conversion is already selected, the method is MKKKKMK M+K+K+K+K+M +K=3.7sec By the requirements document, these two cases are equally likely. Thus, the average time it will take Hal to use this interface for one conver— sion task will be (7.15 + 3.7) / 2 a: 5.4 seconds. But, because the two meth— ods that Hal has to use are different, it will be difficult for him to operate this interface automatically. One of the open problems in the quantitative analy— sis of interfaces is how to estimate error rates from a given interface design. Next, we explore a graphical interface that makes extensive use of a familiar metaphor. 4—2-3-2 Hal’s Interface: Solution 2, GUI The interface shown in Figure 4.3 uses realistic representations of thermometers to indicate temperature. Hal can lower or raise the pointer on each thermometer in Figure 4.3 by using the drag method with the GID. Hal indicates which conversion he wants by moving the arrow on either the Celsius or the Fahrenheit thermometer. He does not type any characters; he simply selects the temperature on. the input thermometer. As he moves one of the pointers, the pointer on the other thermometer moves to the corre— sponding temperature. To set the required precision, Hal expands and con— tracts the scales; he can also change the range. When Hal changes the scale or the range on one thermometer, those on the other thermometer change automatically to cover approximately the same Set of temperatures. Numeriw GOMS KEYSTROKE-LEVEL MODE 81 Figure 4.3. A GUI for Hal’s interface. (See color insert.) 631 readouts are provided on the movable arrow. The temperature is indi- cated both numerically and with a bar, so Hal can use either the graphical or the Character—based representations of the data to accommodate his learning Style or personal preferences. The AutouMed feature changes the ranges such that they are centered on 37 degrees Celsius and 98.6 degrees Fahrenheit, in C356 someone in the lab is working with human body temperatures; this fea— ture is designed to save time. 82 QLJANTIFICATION Clicking on Expand Scales or Compress Scales increases or decreases by a factor of 10 the values at tick marks on the vertical thermometers. To get quickly to a far—distant temperature, Hal expands the scale and scrolls up or down until the desired range is in View, puts the arrow near the desired temperature, and then compresses the scale, adjusting the arrow if necessary, until the desired precision is attained. A GOMS keystroke—level analysis of this graphical interface is com- plex because the method Hal uses depends on where the converter is presently set and what range and precision Hal needs. We look first at the fastest case, in which the range and the precision of the C or the F ther— mometer happen to be already set as Hal wants them to be. This analysis will give us the minimum time needed to use this interface. 0 Write down the gestures Hal uses as he moves his hand to the GID and clicks and holds down the GID button on the desired 31' row: HPK ' Continue listing gestures as Hal moves the arrow until it points to the correct value and then releases the GID button: HPKPK ' Place Ms according to rule 0: HM PMKMK ' Eliminate two Ms according to rule 1: HM PK K There are no cognitive units, no consecutive terminators, and no other rea— sons to apply rules 2 through 5. We find the total time by adding the times for each gesture: H +M + P + K +K 0.4 + 1.35 +1.1 + 0.2 + 0.2 = 3.25 seconds This calculation applies to the lucky case in which the input ther- mometer was preset to the appropriate range and resolution. Now consider the case in which Hal wants to expand the scale factor so that he can see the desired temperature, change the range, compress the scale factor to get ade— quate resolution, and then move the arrow. 1 will write down the method Hal uses, without going through a step—by—step derivation. (I assume that Ha] is a perfect user and does not have to juggle back and forth to find the right places on the thermometer.) Hal has to use the arrows to scroll several times. Each scrolling operation may require several gestures; the computer MEASUREMENT OF INTERFACE EFFlClENCY 83 then animates the scrolling operation, which takes time. To estimate scrolling times for the analysis, I built a similar interface and measured scrolling times which were all 3 seconds or longer. Using 8 to represent the scrolling times1 we can write the sequence of gestures that Hal uses as follows: , HPKSKPKSKPKSKPKK Using the rules to place M5, we get I—I+3(M+P+K+5+K)+M+P+K+K 0.4 + 3 * (1.35 + 0.2 + 3.0 + 0.2) +1.35 + 0.4 + 0.2 + 0.2 = 16.8 seconds Except for the rare case in which the thermometer scales are correctly set at the beginning of the problem, a perfect user will need more than 16 seconds to accomphsh a temperature conversion using this method. A real—imper— fectiuser would jog the scales and the arrows back and forth and thus take even longer. 4-3 Measurement of Interface Efficiency Every tool carries with it the spirit by which it has been created. —Werner Karl Heisenberg We have looked at two interfaces, one of which will take about 5 seconds to operate and the other of which will take more than 15 seconds to operate. It is clear which of the two better satisfies the requirement. The next question that We ask is how fast an interface that satisfies the require— ment can be. Given a design for an interface, you can use GOMS and its extensions to calculate how long a user will take to accomplish any well—defined task with that 1nterface. But analysis models do not answer the question of just how fast You should expect an interface to be. To answer this question, we can use a measure from information theory In the following discussion, information is Esed 1n the technical sense of a quantification of the amount of data conveyed 0': ‘2; 1(ionimumcation, such as when two people have a telephone conversation en a human sends a message, such as a click of the GID button when the guts-or is at a certain location, to a machine. Before dealing with the technical 6129.115 of measuring the amount of information a user must provide to a(2Comphsh a task, we establish the need for such a measurement. 1 To make a reasonable estimate of the time that the fastest possible n tetface for a task would take, we can prOceed by first determining a lower 84 QUANTIFICATION bound on the amount of information a user has to provide to complete that task; this minimal amount is independent of the design of the interface. If the methods of a proposed interface require an input of information that exceeds the calculated lower bound, the user is doing unnecessary work, and the proposed interface can be improved. On the other hand, if the proposed interface requires the user to supply exactly the amount of information that the task requires, you cannot make a more information—efficient interface for this task. In this latter case, there may yet be ways of improving—and there are certainly many ways of ruining—the interface, but at least this one efficiency goal will have been met. Information-theoretic efficiency is defined similarly to the way efficiency is defined in thermodynamics; in thermodynamics we calculate efficiency by dividing the power coming out of a process by the power going into that process. If, during a certain time interval, an electrical gener— ator is producing 820 watts while it is driven by an engine that has an output of 1,000 watts, it has an efficiency of 820 / 1,000, or 0.82. Efficiency is also often expressed as a percentage; in this case, the generator has an efficiency of 82 percent. A perfect generatorfwhich by the second law of thermody— namics cannot exist;would have an efficiency of 100 percent. The information efficiency E of an interface is defined as the minimum amount of information necessary to do a task, divided by the amount of information that has to be supplied by the user. As is true of physical efficiency, E is at least 0 and is at most 1. Where no work is required for a task and no work is done, the efficiency is defined as 1. (This formality is necessary to avoid the case of 0 divided by 0, as in responding to a trans— parent error message. See Section 5—5.) B can be 0 when the user is required to provide information that is totally unnecessary (Figure 4.4). Surprisingly, a number of interface details achieve the dubious honor of having E = 0. A dialog box that allows the user only one possible action, such as clicking the box’s OK button, is such an example. (JavaScript has a command, Alert, solely for creating such unneces— MEASUREMENT OF INTERFACE EFFICIENCY 85 sary boxes: The designers were wise enough to remove goto fiom the javaScript language to force structured code, but they failed to provide simi— lar guidance on the interface side.) E takes into account only the information required by the task and that supplied by the user. Two or more methods may have the same E, yet have dif— ferent total times. It is even possible that a first method has a higher E yet is slower than a second methodefor example, M K MK versus M K K K. In this example, only two characters have to be entered when the first method is used. In the second method, three characters are required, yet it takes less time to perform the task. It is difficult to construct many real—life situations that ex— hibit this inversion of speed and information efficiency.2 For the most part, the more efficient interface is also the more productive, more humane interface. Information is measured in bits; a single bit, which represents a choice between two alternatives—such as 0 and 1, on and off, or yes and no——is the unit of information} For example, a choice made among four objects would require 2 bits of information: If the objects are A, B, C, and D, the first bit could choose either A and B, or C and D; once that choice was made—say C and D—the second bit would choose either C or D. Two binary choices, or 2 bits, suffice to separate one item from a set of four. To choose among eight alternatives, you need 3 bits; to choose among sixteen items, you need 4 bits; and so on. In general, given :1 equally likely alterna— tlves, :he amount of information communicated by all of them taken together is the power of 2 equal to n: log2 :4 And the amount of information in any one of them is (1/n) log2 rt (1) If the probabilities among the alternatives are not necessarily equal and the ith alternative has probability p(i), the information associated with that alter— native is 19(1) 10%;. (1 / PO?) (2) The amount of information is the sum (over all alternatives) of expression (2), which reduces to expression (1) in the equiprobable case. It 2. It is possible to design more sophisticated measures of efficiency; for example, the M oper— ator does not enter into our calculation. However, the simple measure defined here suffices for the purposes of this book. -. 'IS matheniaticmn 50111] Tukey 5 C011 1'3 ] y 5 Bit ) I (I 0 (If i C WI) (IS [HUB dlglll annon and 86 CLUANTlFICATION follows that the information content of an interface that allows only the tap of a single button is 0 bits; not tapping the button is not permitted: 1 log2 (1) = 0 (3) It would seem, however, that the required tap of a single button can, for example, cause the ignition of dynamite used to demolish a building. Would this tap of the button then convey information? It would not, because not tapping the button was not an alternative; the interface “allows only the tap of a single button." If, however, the button was not tapped dur— ing, say, a five—minute time window in which the demoli‘tion was permitted, the building would not be demolished, and the tap or nontap would convey up to I bit of information because there were, in this case, two possible mes— sages. From expression (2), we know that the calculation involves the proba— bility, p, that the building will be exploded. The probability that it will not be exploded is therefore 1 — p. From expression (2), we can calculate the information content of this interface: Plogg (1 /P) + (l #010ng / (110)) (4) When p 2 )4, expression 4 evaluates to ZX1+ZX1=H+M=1 Expression (4) evaluates to less than 1 if p at h”. In particular, it evaluates to 0 when p = 0 or p = 1, as in expression (3). This example illustrates an important point: We can measure the information embodied in a message only in the context of the set of possible messages that might have been received. To calculate the amount of infor— mation that has been conveyed by the reception of a message, we must know, in particular, the probability of that message having been sent. The amount of information in any message is independent of other messages past or future, is without reference to time or duration, and does not depend on any other events; similarly, the outcome of the flip of a fair coin is unaf— fected by previous tosses or by what time of day it is tossed. As explained in Shannon and Weaver (1963), it is also important to keep in mind that information should not be confused with meaning . . . information is a measure of one’s freedom of choice when one selects a message. . . . Note that it is misleading (although often convenient) to say that one or the other message [when just two are possible] conveys [1 bit of] information. The concept of information applies MEASUREMENT OF INTERFACE EFFICIENCY 87 not to the individual messages (as the concept of meaning would), but rather to the situation as a whole, the unit information indicating that in this situation one has an amount of freedom of choice in selecting a message, which it is convenient to regard as a standard or unit amount. (p. 9) However, a user’s actions in performing a task could be modeled with greater accuracy as a Markoff process, whereby the probability of a later action depends on earlier actions taken by the user, but the single_event probabilities discussed are sufficient for the purposes of this book; messages are assumed to be independent and equiprobable. The amount of information conveyed by nonkeyboard devices can also be calculated. If your display is divided into two regions—one labeled Yes and the other labeled Noia single click in one or the other region would supply 1 bit of information. If there are n equally likely targets, with one click, you supply log2 n bits of information. If the targets are of unequal size, the amount of information given by each does not change, but it does take longer to move the GID to smaller targets—by an amount that we shall show how to calculate presently. If the targets have unequal probability, the formula is the same as that already given for keyboard inputs with unequal probabilities. There is a difference in that a user can operate a keyboard key in 0.2 sec, whereas it will take 1.3 sec to operate an on—screen button, on average, ignoring homing time. For our purposes, we can calculate the information content of voice input by treating speech as a sequence of input symbols, rather than as a con— tinuous phenomenon with a certain bandwidth and duration. This treatment of information theory and its relationship to interface design is a simplified account. Yet even in this rudimentary form, informa— tion theory—used in a manner analogous to our use of the simplified QOMS keystroke—level model—can give us firstworder guidance in evaluat~ 1mg the quality of our interface designs. 4-3-1 Efficiency of Hal’s Interfaces Men [oven ofpropre kynde newgflmgelnesse. RChcmcer, "The Squire's File” It is useful to go through a detailed example of a calculation of the average amount of information required for an interface technique. I will agaln use the temperature-conversion example. According to the requirement. 88 OiJANTIFlCATlON the input needed by the converter consists of an average of four typed char— acters; a decimal point occurs once in 90 percent of the inputs and not at all in the other 10 percent, and the negative sign occurs once in 25 percent of the inputs and not at all in the other 75 percent. For simplicity, and because there is no need for 1 percent precision in the answer, I will assume that all of the other digits occur with equal frequency, and I will ignore the 10 percent of the inputs that have no decimal point. We need to determine the set of possible messages and the probabil— ity of each. Five forms are possible, where d denotes a digit: 1. —.da' 2. —d.d 3. .ddd 4. d.dd and 5. dd.d. The first two each occur 12.5 percent of the time, and there are 100 of each of them; the final three each occur 25 percent of the time, and there are nearly 1,000 of each.4 The probability for either of the first two types of messages is (0.125 / 100) = 0.00125; the probability for any one of the final three types of messages is (0.75 / 3000) = 0.00025. The sum of the proba— bilities of the messages is, as it must be, 1. The amount of information of each message, in bits, is given by expression (2)5: p0) Iota (1 / at?) This expression evaluates to approximately 0.012 for the negative values and to 0.003 for the positive values. Calculating 200 X 0.0067 + 3000 X 0.003 gives a total of 11.4 bits for each message. Taking the probabilities into account can be important. If we took a simple—minded approach and assumed that all of the 12 symbols (minus, decimal point, and the 10 digits) were equally likely, the probability of each would be X2, and the information contained in a four—character message would be approximately 410g2 (12) a 14 bits 4. The “nearly” comes from the fact that the temperature of 0 degrees will not be entered as 0.00 01' 00.0. 5. To get logs to the base 2 on a calculator or a computer that has only natural logs (In), use: log2 (x) = ln / ln MEASUREMENT OF INTERFACE EFFICIENCY 89 It is a theorem of information theory that the information is at a maximum when all symbols are equally likely. Therefore, making the assumption of equiprobable messages will give you a value that is equal to or greater than the amount of information in each message. Obviously, this assumption also makes estimating the information content of a message eas— ier to compute. If the resultant value of the approximation is smaller than the amount of information your interface requires the user to supply, you do not yet need to bother with the more refined calculation. We have just calculated that the task requires that Hal supply an aver— age of about 11 bits of information each time he has to convert a tempera— ture. We can—and will, presentlyeedivide this quantity by the amount of information the interface requires him to supply. The result will be the effi— ciency of the interface. Another simplification for quick analysis is to find the amount of information in a keystroke or a GID operation and then to count the various gestures. When a keystroke delivers information to a computer, the amount of information delivered depends on the total number of keys available—for example, the number of keys on the keyboard—and the relative frequency with which each key is used. Thus, keystrokes can be used as a rough mea— sure of information. If a keyboard had 128 keys, each of which had the same frequency of use, each key would represent 7' bits of information. In prac— tice, the frequency of use varies tremendously—for example, space and e are common, whereas j and \ are rare), and the information per keystroke is closer to 5 bits in most applications. The requirement stated that the average length of the input that specifies the temperature was four keystrokes. For this analysis, it is easier to use a measure simpler than information— theoretic efficiency but that often achieves the same practical effect. Charac- ter efficiency is defined as the minimum number of charaCters required for a task, divided by the number of characters the interface makes the user enter. Achieving an interface that required four keystrokes, on average, would give us a character efficiency of 100 percent. If we add a keystroke to decide which conversion is desired and then another to delimit the answer, our average length of input will grow to six keystrokes, and our keystroke efficiency will drop to 67 percent. If Hal has as his input device only a 16— key numeric keypad, the information provided by a single keystroke would be 4 bits, and the interface would be more efficient. (The requirements, however, do not permit us to use this option.) Because any task in a GOMS analysis requires at least one mental opera ator, the most keystroke—efficient interface for the temperature—conversion problem will have, in theory, an average time of 90 QIJANTIFICATION M+K+K+K+K=2.155ec Thus, it will be considerably faster than either of the two interfaces already discussed. However, typing four characters on a standard keyboard supplies at least 20 bits of information, whereas only 10 bits are required—an information—theoretic efficiency of 50 percent—so we know that there is room for improvement. As we have seen, using a standard numeric keypad instead of a full keyboard drops the input information per four keystrokes to 16 bits, raising the efficiency to 62 percent. A dedicated numeric keypad— one that has only the digits, the minus sign, and a decimal pointiwill per— mit a slightly higher score, of about 70 percent efficiency. We raise the score again by using special encodings of temperature information and novel input devices, but training difficulties and excessive costs begin to loom with these extreme approaches, so I will stop here and accept 70 percent information— theoretic efficiency. Theoretical limits may or may not be reached by a prac— tical interface, but they do give us a star by which to steer. 4-3-2 Other Solutions for Hal’s Interface In Section 4—3—1, we stopped trying to improve information— theoretic efficiency when we reached 70 percent. We achieved that effi— ciency with an unspecified, theoretical interface that somehow managed to have 100 percent keystroke efficiency. Let us see how close we can come to this ideal with a standard keyboard and a GID. Consider an all—keyboard interface. In this interface, a note appears on the display: To convert temperatures, indicate the desired cale by typing C for Celsius or F for Fahrenheit. Type the numeric tempera— ture; then press the Enter key. The con- verted temperature value will be displayed. A GOMS analysis finds that the user must make six keystrokes. Following the rules for placements of Ms gives us MKKKKKMK The average time is 3.9 seconds. MEASUREMENT OF INTERFACE EFFICIENCY 91 We can decrease this time if we can use the C or the F itself as a delimiter. That is, consider an interface in which the following instructions appear: To convert temperatures, type the numeric temperature, followed by C if it is in degrees Celsius or F if it is in degrees Fahrenheit. The converted temperature will be displayed. In this example, the Enter key is not used. Some primitive interface—building tools demand that the user tap Enter and will not permit us to use C or F as a delimiter; such tools are inadequate for building humane interfaces. The GOMS analysis of the C/F—delimiter interface yields MKKKKMK The average time is 3.7 seconds. If we did not have an analysis that showed that the theoretical minimum time is 2.15 sec, this solution might strike us as satisfactory. It is considerably more efficient than the ones that we discussed previously, so we might stop here. Tempted by that theoretical minimum, however, we ask whether there is an even faster approach. Consider the interface depicted in Figure 4.5; we might describe it as bifurcated: One input will give us two outputs. Temiiééiiature tonuerter Type in the temperature to be converted. The converted temperature will appear on the right as you type. Figure 4.5. An interface that does not require a delimiter. A more effi— cient interface is made possible by taking advantage of character-atm- time interaction, and by performing both conversions at once. 92 QUANTIFICATEON Under the bifurcated interface, no delimiter is reqired. Further— more, the user does not have to specify which conversion; desired. The GOMS analysis for the average input of four characters is MKKKK The bifiircated interface achieves the minimum 2.15 secons and has 100 percent character efficiency. If, as in our example, the output sometimes changes wen a character is typed, the flickering of the output does not distract you because your locus of attention is the input. The continually changing utput is often beneficial: The user will notice it only peripherally aftler theirst few times that he uses the feature, at which point it will provide him fedback that the system is responding to his input. For single—character intraction to be effective, the system must respond quickly; in particular, 1e interaction must keep up with the user’s typing speed. Only a slow neter connection should exhibit this problem. Although not part of the requirement, you might ask hO‘ this converter is “cleared” for the next operation. Does the clear operation ad a keystroke? Not necessarily. For example, we could design the interface suchhat, whenever the operator returns to his background task or goes on to anotlr task, the val— ues in the converter are automatically grayed and the convertenecomes inac— tive. The values shown are not cleared at this time, so that they ca be referred to again if necessary. The next input to the converter does clear thold values. just because it has optimal speed of operation and is lghly efficient, the bifurcated converter is not necessarily the best interface of 1056 discussed or of those possible. Parameters other than speed also are of imommce1 such as error rate, user learning time, and long—term user retentiorof the way to use the interface. We should be especially concerned about tl error rate of the bifurcated converter, due to Hal’s possibly reading the wr0g output box, especially because he may have just heard, for example, the Wrd Celsius and thus be required to read out the Fahrenheit line. Nonethelesshe bifurcated converter would definitely be on the short list of interfaces the tested for the temperatureuconverter application, and a few others that v. have seen— solutions that might otherwise have seemed worth a try had is not learned how to do a GOMS analysis—would not make the cut. Whether we use it in a simple keystroke—timing arlysis or in a detailed informationetheoretic extravaganza, a quantification Cthe theoreti— cal minimum—time, minimum—character, or minimum—inforrrrion interface can be a useful guide for our designs. Without a quantitative grit; we are only guessing at how well we are doing and at how much room there is for iprovement. FITTS’ LAW AND l—llCK'S LAW 93 4-4 Fitts’ Law and Hick’s Law It behooves its [opiate thejoundations of/eiiowiedge in mathematics. iRoger Bacon, Opus Majus (13th century) Various quantitative laws relating to interface design have sound cog- netiC underpinnings and have been validated repeatedly. These laws often give you additional data on which you can base interface—design decisions. Fitts’ law quantifies the fact that the farther a target is fi:0m your current cursor posi— tion or the smaller the target is, the longer it will take you to move the cursor to the target. Hick’s law quantifies the observation that the more choices of a given kind you have, the longer it takes you to come to a decision. 4—4—1 Fitts’ Law Consider that you are moving a cursor toward an on—screen button. The button is the target of the move. The length of a straight line from the position at which the cursor started to the closest point on the target is the distance used in the statement of Fitts’ law. Given the size of the target and the distance to be moved, Fitts’ law gives you the average time it takes a user to succeed in. getting the cursor to the button. In the one—dimensional case, in which the target’s size, measured along the line of motion, is S and the target is at a distance D from the start“ ing position (Figure 4.6), Fitts’ law states that Time (in msec) = a + l) log2 (D / S + l.) (The constants a and b are determined experimentally or are derived from human performance parameters.)6 The time that you calculate begins when 6. Mathematics, supposedly a paragon of clarity, clings yet to that old-fashioned style whereby undefined variables appears in a formula before you know what they stand for. For example, you will see such statements as A : 'IT r2, where ris the radius of a circle and A is its area. This can be confusing, forcing you to read ahead and then go back, especially if the equation is a long one with lots of as-yet—unexplained variables. Far better, from a reader’s viewpoint, is to follow the obvious dictum to define terms befiire you use them: A circle with radius r has an area A, given by: A=Ttr2 94 QUANTIFICATION initial Cursor I‘— 5 —’l Position i«—o—» Figure 4.6. Distances used in Fitts’ law to determine the time to move a cursor to a target. the cursor is at the starting point and after the user has chosen the target. The logarithm to the base 2 gives a measure of the difficulty bf the task in terms of the number of bits of information it takes to describe the (one—dimensional) path of the cursor. The units of distance do not affect the calculated time, because D / S is the ratio of two distances and is therefore dimensionless. It follows that, even though we might move the pointing device a distance smaller or larger than the distance the cursor moves on the display, the law still works when the distances are measured on the display, assuming a linear relationship between GID and Cursor motion. Fitts’ law applies only to the kinds of motions we make when we are using most human-machine interfaces: motions that are small relative to human body size and that are uninterrupted, that is, move— ments that can be made in one continuous motion. For back—of—the—envelope approximations, I use a = 50 and b = 150 in the Fitts’ law equation. An extension of Fitts’ law to more complex constraints, such as tracking a cursor between straight or curved walls, has been developed and tested empirically (Accot and Zhai 1997). For a two—dimensional target, you can usually obtain a reasonable approximation of the time needed to move the cursor to the target, using the smaller of the horizontal and vertical dimensions of the target for the value of S (Mackenzie 1995). Fitts’ law explains, for example, why it is much faster to move the cursor to an Apple Macintosh—style menu (Figure 4.7) that is on the edge of a display than to a Microsoft Windows—style menu (Figure 4.8) that floats away from an edge. The size S of the Windows menu on my display is 5 mm. The effective size of the Macintosh target is large because you do not have to stop within the confines of the menu bar but rather can continue to move the GlD any comfortable distance beyond that needed to put the cur— sor in the menu: The cursor stops at the edge of the display. A series of tests I performed determined that users typically stop Within about 50 mm of the edge of the display on the Macintosh, so we can use 50 min as S for the Macintosh. On a 14—inch flat panel display, the aver— FITTS' LAW AND HICK'S LAW 95 Figure 4. 7. The Macintosh menu, at the top edge of the screen, e;ka- tioely increases its size compared to a menu that floats away from the edge. (See color insert.) «a, America I] Figure 4.8. The Windows menu is below the top edge of the screen; you have to place the cursor more carefiilly to pull down a submenu. (See color insert.) age distance the cursor must be moved to reach the menu bars is 80 mm; thus, the calculated time to move the cursor to a menu item on the Macin— tosh is 50 + 15010g2 (80 / 50 + 1) = 256 msec. This reSult is far less than the calculated time it takes to move the cursor to a corresponding menu item on a Windows—style menu: 50 + 150 log2 (80 / 5 + 1) = 663 msec. These calculations apply only to the time it takes the user to move the cursor. Clicking on the target to indicate that you believe that the cursor has reached its goal adds another 0.1 sec on average. (K = 0.2 in the GOMS model includes both the downstroke and the release of the button, whereas this timing is stopped by the downstroke.) In a typical experimental situa— tion, you have to add the human reaction time of about 0.25 sec at the start 96 CLLJANTIFICATION of the cursor movement. When we ake these factors into account, we get times that agree with what I have oherved: It takes about 0.6 sec, on aver- age, for a user to open an Apple melu, whereas it takes a user more than 1 sec to open a Windows menu. This aialysis makes it clear why menus were deliberately placed at the edge of thfdisplay when the Macintosh interface was developed. 4-4—2 Hick’s Law Before you move the cursor to a targ-t or take any one action with a multi— plicity of choices, you must first chose the target or attion. Hick’s law says that when you have to choose to talc: one among rt alternative actions and when the probabilities of taking eah alternative are equal, the time to choose one of them is proportionalto the logarithm to the base 2 of the number of choices, plus 1. When pit this way, Hick’s law looks just like Fitts’ law: Time (in msec) = a + blogg (H I) If the probability of the ith choice is 19(1), then, instead of the logarithmic factor in the equation, you use ape Iogz (1 / pa) + i) The coefficients in Hick’s laware strongly dependent on many con— ditions, including how the choices at- presented and how habituated to the system the user has become. (If thechoices are presented in a confusing manner, both a and b can increase; hbituation decreases in.) These depen— dencies Will not he discussed here; i we need to consider is that making decisions takes time, that making couplex decisions takes more time than making simple ones, and that the relaionship is logarithmic. In the absence of better information, we can use thisame coefficients a and b as for Fitts’ law to make off—the—cuff or relative Cfiimates, Whatever positive, nonzero CDfficients we use for a and b, it follows from Hick’s law that giving a user flmy choices simultaneously is usually faster than is organizing the same chdces into hierarchical groups. Making choices from one menu of eight item is faster than is making choices fiom two menus of four items each. Assurrng that the items are equally liker to be chosenmand ignoring the time it tkes to open the second menu, which, if taken into account, would make th time taken for the two—menu inter— face even longer—we compare the the to select one item of eight, a + b FITTS' LAW AND HICK'S LAW 97 logz 8, with the time to select one item of four twice, 2 (a + b log2 4). Because log2 8 = 3 and log2 4 = 2, and because both a < 2a and 3?) < 45, we see that a+3b<2(a+2b) This accords with experiments on menu structures (see, for example, Norman and Chin 1988). ‘ Our discussion of Fitts’ and Hick’s laws is incomplete, For example, it is no accident that they have the same form as the Shannon—Hartley theo— rem. Nonetheless, this brief treatment is sufficient to alert you to these use— ful guides to interface design. They can help you even if, as in the example, you do not know the empirical coefficients a and b. (For more detail, see Card, Moran, and Newell 1983, pp. 72—74.) ...
View Full Document

Page1 / 14

raskin_chp3_4 - 70 MEANINGS. MODES. MONOTONY. AND MYTHS...

This preview shows document pages 1 - 14. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online