computer/human interface design

  • Upload
    lh

  • View
    216

  • Download
    0

Embed Size (px)

Citation preview

  • 8/15/2019 computer/human interface design

    1/295

    The Psychology of Menu

    Selection: Designing

    Cognitive Control at theHuman/Computer

    Interface

     by Kent L. Norman 

     published by Ablex Publishing Corporation, 1991, 36 pages!"#$N% &'9391'((3')*.

    About the oo! 

    +enu seletion is emerging as an important mode o- humanomputer interation. /his boo0, the -irst entirely deoted to this important -orm o- humanomputer interation, proides detailed theoretial and empirial in-ormation o- interest to so-t2are designersand humanomputer interation speialists and researhers. A ne2 theoretial approahto menu seletion is ta0en by deeloping a psyhologial theory o- ognitie ontrol bythe user. A omprehensie reie2 o- empirial researh on menu seletion is presentedin an organied -ashion to aid in the design and ealuation o- systems. 4inally,in-ormation is gien on ho2 to protype and ealuate menu seletion systems using both

     per-ormane data and user ratings.

    /he olume has three parts. Part 5ne is oneptual and theoretial in nature. /he -irsthapter introdues the issues o- design and -lo2 o- ontrol at the humanomputerinter-ae. "n the next three hapters taxonomi -rame2or0s are proposed onerning the

    type o- menu seletion system being used, the nature o- the tas0 being per-ormed by theuser, and the ognitie elements inoled in per-orming the tas0. "n Part /2o,experimental researh on menu seletion stemming -rom paradigms deeloped inexperimental psyhology and more reently human -ators and ognitie psyhology isdisussed. /he last part o- the boo0 deals 2ith the topi o- implementation andealuation. Chapters disuss priniples o- 2hen and ho2 to use menus, oer topis o-

     prototyping and ealuation, and attempt to plot some o- the -uture diretions o- menuseletion. /hroughout, graphs and illustrations are inluded. xamples o- good and baddesigns are sho2n in a number o- illustrations 2hile empirial data -rom experimentsare desplayed in graphs.

    /he reader 2ill bene-it -rom the disussion o- the many issues, design possibilities andinsights regarding menu sletion. /he empirial researh at times supports and at other

    http://www.lap.umd.edu/lap/people/kent_norman/index.htmlhttp://www.lap.umd.edu/lap/people/kent_norman/index.html

  • 8/15/2019 computer/human interface design

    2/295

    times re-utes existing guidelines. /he reader 2ill 2ant to 0no2 2hat the urrent state o-0no2ledge is about ho2 to design menuy seletions and 2hy the design hoies areimportant.

    The "n#$ine Copy

    /he omplete table o- ontents is listed belo2. 7ou may hae a pee0 at the boo0 byli0ing on the hapter titles. 7ou may use this resoure -reely8 ho2eer, " 2ould pre-erthat you atually buy a opy to help de-ray publishing osts. Also, " should point outthat the on'line opy is -rom a pre'publiation dra-t and has not been are-ully proo-read -or errors.

    Table of Contents

    Preface

    Part I: The Theory an% Implementation of Menu Selection Systems

    1. "ntrodution to the /heory o- Control at the umanComputer "nter-ae 1.1 :esearh and ;esign o- the umanComputer "nter-ae 1.1.1 "ssues in ;esign

    1.1.< /hree Paradigms o- ;esign1.< A +odel o- the umanComputer "nter-ae 1.

  • 8/15/2019 computer/human interface design

    3/295

    .> "nput% ;ata or Parameter #pei-iation3.( 5peration by +enu #eletion% Command +enus 

    3.6 ;eision by +enu #eletion% ;eision +enus 3.6.1 ;eision /rees3.6.< ;eision +atries3.B "n-ormation :etrieal by +enu #eletion% "n-ormation +enus 3. Classi-iation by +enu #eletion% Category +enus 3.9 #ummary >. Cognitie lements o- +enu #eletion >.1 /he +enu #eletion Proess >.1.1 "n-ormation A?uisition and #earh>.1.< Choie Proess and /ime>.1.3 :esponse Proess>.1.> aluation and rror ;etetion>.< Problem #oling and #earh #trategies >...3 Cognitie Layouts o- +ental +odels >.3.1 +enu #eletion as a +etaphor>.3.< #hemata and #ripts>.3.3 Cognitie Layouts o- +enus>.> #ummary 

    Part II: Design &ui%elines from 'mpirical (esearch(. :esearh "ssues and +ethods in +enu #eletion (.1 "ntuition and ;ata in Con-lit (.< :epliability (.3 "mportane o- the :esult (.> @eneraliation o- :esults (.( xperimental ;esigns (.6 #ummary 6. 4ormatting and Phrasing the +enu 6.1 4ormatting the +enu 4rame 6.1.1 Amount o- "n-ormation per #reen

    6.1.< 4ousing Attention on the +enu6.1.3 Pereptual @rouping

    http://www.lap.umd.edu/poms/chapter2/chapter2.html#2.3http://www.lap.umd.edu/poms/chapter2/chapter2.html#2.4http://www.lap.umd.edu/poms/chapter3/chapter3.htmlhttp://www.lap.umd.edu/poms/chapter3/chapter3.html#3.1http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.2http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.3http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.4http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.5http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.6http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.7http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.7http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.8http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.9http://www.lap.umd.edu/poms/chapter4/chapter4.htmlhttp://www.lap.umd.edu/poms/chapter4/chapter4.html#4.1http://www.lap.umd.edu/poms/chapter4/chapter4.html#4.2http://www.lap.umd.edu/poms/chapter4/chapter4.html#4.3http://www.lap.umd.edu/poms/chapter4/chapter4.html#4.4http://www.lap.umd.edu/poms/chapter5/chapter5.htmlhttp://www.lap.umd.edu/poms/chapter5/chapter5.html#5.1http://www.lap.umd.edu/poms/chapter5/chapter5.html#5.2http://www.lap.umd.edu/poms/chapter5/chapter5.html#5.3http://www.lap.umd.edu/poms/chapter5/chapter5.html#5.4http://www.lap.umd.edu/poms/chapter5/chapter5.html#5.5http://www.lap.umd.edu/index.htmlhttp://www.lap.umd.edu/poms/chapter6/chapter6.htmlhttp://www.lap.umd.edu/poms/chapter6/chapter6.html#6.1http://www.lap.umd.edu/poms/chapter2/chapter2.html#2.3http://www.lap.umd.edu/poms/chapter2/chapter2.html#2.4http://www.lap.umd.edu/poms/chapter3/chapter3.htmlhttp://www.lap.umd.edu/poms/chapter3/chapter3.html#3.1http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.2http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.3http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.4http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.5http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.6http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.7http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.8http://www.lap.umd.edu/poms/chapter3/chapter3.html#3.9http://www.lap.umd.edu/poms/chapter4/chapter4.htmlhttp://www.lap.umd.edu/poms/chapter4/chapter4.html#4.1http://www.lap.umd.edu/poms/chapter4/chapter4.html#4.2http://www.lap.umd.edu/poms/chapter4/chapter4.html#4.3http://www.lap.umd.edu/poms/chapter4/chapter4.html#4.4http://www.lap.umd.edu/poms/chapter5/chapter5.htmlhttp://www.lap.umd.edu/poms/chapter5/chapter5.html#5.1http://www.lap.umd.edu/poms/chapter5/chapter5.html#5.2http://www.lap.umd.edu/poms/chapter5/chapter5.html#5.3http://www.lap.umd.edu/poms/chapter5/chapter5.html#5.4http://www.lap.umd.edu/poms/chapter5/chapter5.html#5.5http://www.lap.umd.edu/index.htmlhttp://www.lap.umd.edu/poms/chapter6/chapter6.htmlhttp://www.lap.umd.edu/poms/chapter6/chapter6.html#6.1

  • 8/15/2019 computer/human interface design

    4/295

  • 8/15/2019 computer/human interface design

    5/295

  • 8/15/2019 computer/human interface design

    6/295

    Chapter ) 

    Intro%uction to the Theory of Control at the Human/Computer Interface 

    "n the deelopment o- any ne2 tehnology, the last -rontier al2ays seems to be the

    enounter 2ith the end user o- the produt. /he hallenge is to deelop a produt thatan be easily and e--iiently used to aomplish its intended purpose. /he 0ey elementis to design an inter-ae bet2een the user and the mahine that allo2s -or the proper-lo2 o- ontrol bet2een human and mahine proesses. /his is partiularly true o-omputer based systems. /he deelopment yle o- omputers is no2 2ell past the

     point 2here the use o- omputers is exlusiely by omputer sientists. /he general2or0 -ore and the publi'at'large is already using or beginning to use the omputer inone -orm or another. "t is at this point that are-ul onsideration must be made o- ho2 tolay out the arhiteture and the building ode in eah ne2 ommunity o- users. ithoutsuh -orethought, users are li0ely to beome prematurely lo0ed into an arbitrary andarhai struture.

    Among the ne2 arhitetures o- the humanomputer inter-ae that spei-ially deal2ith -lo2 o- ontrol is the design o- menu seletion systems. 4igure 1.1 illustratesseeral suh systems. =sers are presented 2ith a list o- options -rom 2hih they anhoose and some mehanism by 2hih to indiate their hoie. /he harateristis o-menu seletion are that !a* the interation is, in part, guided by the omputer8 !b* theuser does not hae to reall ommands -rom memory, and !* user response input isgenerally straight -or2ard. /he -our examples sho2n in 4igure 1.1 highlight the arietyo- suh menu systems.

  • 8/15/2019 computer/human interface design

    7/295

  • 8/15/2019 computer/human interface design

    8/295

    +enu seletion is emphasied in this boo0 as a priniple mode o- ontrol used inonFuntion 2ith other modes suh as -orm -ill'in, ommand languages, naturallanguage, and diret manipulation. "t is -elt by the author that in terms o- -lo2 o-ontrol, menu seletion is emerging as the mode o- hoie. 5ther modes ome into playto handle di--erent demands on the humanomputer inter-ae. #2ithing -rom one

    mode to another is o-ten neessary and must be done grae-ully 2ith lear expetationon the part o- the user. #peialied modes, their integration 2ith menu seletion, and the

     problems o- s2ithing bet2een modes 2ill be disussed as it is appropriate.

    "n this hapter 2e 2ill onsider the humanomputer inter-ae in terms o- -lo2 o-ontrol. /he user has ertain tas0s to aomplish and, onse?uently, 2ants to diret theomputer to per-orm a subset o- those tas0s. /he problem -rom the userGs perspetie is0no2ing 2hat the omputer an do and 0no2ing ho2 to diret it to do those tas0s. /he

     problem -rom the system designerGs point o- ie2 is 0no2ing 2hat -untions toimplement and ho2 the omputer should in-orm the user about the aailability andinoation o- these -untions. "n doing so, the omputer must be assigned a ertain

    measure o- ontrol oer the -lo2 o- operations8 but ultimately, ontrol rests 2ith theuser.

    "n the next setion the proess o- designing the humanomputer inter-ae and itsimportane 2ill be disussed. 4ollo2ing this a ognitie model o- the humanomputerinter-ae is proposed that helps to de-ine and spei-y the issues o- ho2 humans 2or02ith mahines and ho2 mahines must be designed so that they are usable by humans.4inally, the use o- experimental methods in the design proess 2ill be disussed.mpirial researh is the mainstay o- human -ators and ergonomis, and it is importantto understand ho2 and 2hy experiments are onduted.

    )*) (esearch an% Design of the Human/Computer Interface 

    5ne o- the maFor steps in the so-t2are deelopment li-e yle is the translation o-system re?uirements into an integrated set o- tas0 spei-iations that de-ine the-untional apabilities o- the so-t2are and hard2are !Hensen I /onis, 19B9*. /he

     problem is that suh re?uirements are o-ten ditated by analysts 2ith little thought aboutthe end user. Conse?uently, userGs needs and limitations are o-ten not -atored into thee?uation. No matter ho2 2ell implemented other parts o- the system may be, i- thehumanomputer inter-ae is intratable, the system 2ill -alter and ultimately -ail.Ledgard, #inger, and hiteside !191* point out the inredible ost o- poor human

    engineering in the design o- interatie systems. /hey identi-y -our osts%

    1. /he diret osts o- poor design are obsered in 2asted time and exessie errors. /hesystem itsel- may hae been designed -or maximum e--iieny in terms o- memorymanagement, inputoutput, and omputation but, it may sit idly by as the user pondersan error message or thumbs through doumentation loo0ing -or the orret ommand.+enu seletion systems may alleiate suh 2asted time by sel-'doumenting the systemand redue errors by limiting user input to only the set o- legal options.

  • 8/15/2019 computer/human interface design

    9/295

  • 8/15/2019 computer/human interface design

    10/295

    suh ergonomi -ators and the trade'o--s bet2een them depend, o- ourse, on the partiular tas0 and the user ommunity.

    /able 1.1

    4ators to be Considered in the ;esign o- umanComputer #ystems

       

    #ystem Produtiity

    Appliability o- system to tas0

     Number o- tas0s ompleted

    Muality o- output

    uman Per-ormane

    #peed o- per-ormane

    :ate and type o- errors

    Muality o- solutions to problems

    /raining time and e--etieness

    /ime to learn ho2 to use the system

    4re?ueny o- re-erene to doumentation

    uman retention o- ommands oer time

    /rans-er o- training

    Cognitie Proesses

    Appropriateness o- the mental model

    ;egree o- mental e--ort

    #ubFetie satis-ation

    #atis-ation 2ith sel-

    #atis-ation 2ith system

    #atis-ation 2ith per-ormane

  • 8/15/2019 computer/human interface design

    11/295

  • 8/15/2019 computer/human interface design

    12/295

  • 8/15/2019 computer/human interface design

    13/295

    )*)*+ Three Para%igms of Design* "t has been pointed out by Anderson and 5lson!19(*, #terling !19B>*, and olt and #teeson !19BB* that human -ators onsiderationsmust be integrated into the design proess -rom the beginning and as it progresses.=n-ortunately, ery -e2 systems hae been designed 2ith human -ators as a high

     priority item. /he reason -or this may be traed to the historial deelopment o-

    mahines, and it 2ill proe instrutie to reie2 this design proess.

    istorially, the design o- manmahine systems proeeded -irst 2ith the po2er unit, thedrie mehanism, the mill, and -inally ontrol as sho2n in the top panel o- 4igure 1.

  • 8/15/2019 computer/human interface design

    14/295

  • 8/15/2019 computer/human interface design

    15/295

    /his paradigm is being employed by more and more o- the industry. /eams o- human-ators speialists are being lin0ed up 2ith so-t2are deelopment teams to proideinitial analysis o- tas0s and user harateristis, to ondut researh on prototypes, andto ondut user aeptane tests on the -inal produt.

    /his approah is partiularly important -or systems that are heaily dependent on the-lo2 o- ontrol at the humanomputer inter-ae. /he 2or0 o- humanomputerinteration speialists in design rests on t2o maFor omponents% theoretial models o-the manmahine inter-ae and methods o- applied experimental researh. /hese aredisussed in the next t2o setions.

    )*+ A Mo%el of the Human/Computer Interface 

    +odels o- the humanomputer inter-ae depend heaily on ognitie psyhology. /he psyhologial proesses o- attention, memory, in-ormation proessing, deision ma0ing,and problem soling must be ta0en into aount. 5ne o- the most important -eatures in

    suh models is the -lo2 and -eedba0 o- in-ormation through the inter-ae. /he userneeds in-ormation -rom the omputer and the omputer annot -untion 2ithoutin-ormation -rom the user. A maFor omponent o- this interation is the -lo2 o- ontrolin-ormation. /he omputer gies in-ormation to prompt the user -or input and the usersupplies input that direts the subse?uent operations. #mooth operation o- the systemre?uires a timely -lo2 o- in-ormation that is relatiely -ree o- error states in the mahineand in the user. rror states in the mahine an be 2ell de-ined. @enerally mahineerrors at the inter-ae our 2hen !a* input alues -all outside an allo2able range ordisagree 2ith re?uired type, !b* re?uired resoures are not aailable, or !* a all ismade to a non'existant -untion or loation. rror states also our in the user. "nontrast to mahine errors, user error states are not 2ell'de-ined sine they arise -romsubFetie states 2hen -or some gien omputer output or prompt, the user does not0no2 2hat to do. =ser error states are harateried by on-usion, la0 o-understanding, and la0 o- 0no2ledge o- 2hat to do next.

    4igure 1.3 sho2s a shemati model o- the -lo2 o- in-ormation and ontrol at thehumanomputer inter-ae that has been adapted -rom Norman, et al. !191*. /he model-irst emphasies that the system is embedded 2ithin a tas0 situation. /he user, -orexample, may be monitoring an industrial proess, or he may be engaged in in-ormationretrieal. /he tas0 determines a number o- oerriding -ators, suh as the ost o- errors,importane o- speed, and onsiderations that de-ine the suess-ul ompletion o- the

    tas0. $oth the human !represented by a irle* and the mahine !represented by aretangle* reside in enironments that proide in-ormation, onstraints, and ontexts. "tmust be remembered that the user interats not only 2ith the mahine but more-undamentally 2ith the enironment. e or she attends to in-ormation oming in andgenerates in-ormation going out. A similar piture is sho2n -or the mahine 2hih maymonitor an enironment and retriee in-ormation as 2ell as generate output to theenironment.

  • 8/15/2019 computer/human interface design

    16/295

  • 8/15/2019 computer/human interface design

    17/295

    ognitie proesses. Although 2e are only beginning to understand the proesses o-mental operations, 2e an obsere the produts o- thought and generate theories aboutthe li0elihood o- partiular responses on the part o- the user. "n humanomputerinteration, -lo2 o- ontrol is shared and at times passed ba0 and -orth bet2een theuser and the mahine. "n light o- our limited 0no2ledge about user thought proesses

    and the added omplexity inurred by shared ontrol, it is no 2onder that the design o-the humanomputer inter-ae is no easy matter and that it remains as a maFor area o-onern in omputer siene, ognitie psyhology, and ergonomis.

    Although the model in its present -orm is largely oneptual in nature, it helps todelineate the onerns that an ergonomis speialist should hae 2hen helping to designthe inter-ae. /hese inlude the -our areas sho2n in 4igure 1.3% !a* the harateristis o-the tas0 and the enironment, !b* the harateristis o- the human ognitie proessing,!* the spei-iations -or the omputer proessing, and !d* the implementation o- thehumanomputer inter-ae.

    )*+*) Characteristics of Tas!s an% 'nvironments* "t is obious that di--erent tas0s anddi--erent enironments impose di--erent needs and onstraints on humanomputerinteration. hat is not so obious is ho2 to meet these re?uirements. An analysis o-the tas0 and enironment is the -irst step. /a0e, -or example, the tas0 o- balaning oneGshe0 boo0. /he tas0 re?uires data input to the omputer, eri-iation o- reords, and aseries o- omputational steps. /he enironment inludes in-ormation about urrent

     balane, aneled he0s, et., as 2ell as time and resoure onstraints that a--et themotiation andor -rustration leel o- the user.

    /as0s may be harateried along the -ollo2ing dimensions%

    Simple--Complex. #imple tas0s inole -e2 steps 2ith little demand on the user or onthe omputer. Complex tas0s inole many steps and impose high demands on the userand possibly the omputer as 2ell. 5- ourse, the tas0 as implemented may alloatesimple or omplex parts to the human or omputer. riting a noel on a 2ord proessor is a omplex tas0 -or the human and a omparatiely straight-or2ard one -or theomputer. 5n the other hand, an in-ormation retrieal problem may be ratherstraight-or2ard to the user but omparatiely omplex -or the system.

    Structured--Unstructured . #trutured tas0s hae a preplanned ourse8 2hereas,unstrutured tas0s may inole reatie planning and rediretion. "n strutured tas0s,

    suh as he0boo0 balaning, the -lo2 o- ontrol may be relegated to the omputer. 5nthe other hand, -or unstrutured tas0s, suh as 2riting a noel, the user maintainsontrol oer most aspets o- the tas0 exept the mundane operations o- omputerhouse0eeping.

    /as0s may be harateried by many other dimensions suh as the degree to 2hih inputs. output predominates or the extent to 2hih the user is an atie s. passie

     partiipant. Chapter 3 2ill disuss tas0 analysis in greater detail8 ho2eer, at this pointit is su--iient to be a2are o- the -at that tas0 harateristis impose ertain demands onthe user and on the system.

    nironments are o-ten lin0ed to tas0s, but 2e may also de-ine t2o harateristis o-enironments as -ollo2s%

  • 8/15/2019 computer/human interface design

    18/295

    Time Critical--Resource Limited . "n many situations the enironment imposes a timeonstraint. 4or example, the in-ormation must be retrieed be-ore the #enate#ubommittee hearing at 1&%&& am8 the deision to deploy the torpedo must be made in3& seonds. #uh enironments hae a psyhologial impat on the user and re?uirethought as to ho2 to implement the humanomputer inter-ae to ahiee aeptable

    leels o- per-ormane.

    Controllable--Immutable. /he user and the system may be able to alter theenironment. /his may be an inherent part o- the tas0 in industrial ontrol. Control o-resoures adds an additional leel o- onern to the user and needs to be onsidered interms o- the ognitie demands on the user. 5n the other hand, the enironment may beimmutable in the sense that neither the user or the omputer an hae an a--et on it.

    nironments may also be harateried by a number o- other dimensions suh as2hether they are in-ormation rih s. in-ormation sare, sa-e s. haardous, et. /heseharateristis 2ill proe to be important 2hen it omes to the design and ealuation o-

    the user inter-ae partiularly as it relates to -lo2 o- ontrol.

    )*+*+ Characteristics of the Human ,ser* /2o ie2s o- the user exist. /he user iseither an extension o- the system or the system is an extension o- the user. "n the -irstase, the user may, un-ortunately, be ie2ed merely as an input deie not unli0e anoptial sanner or an analog'to'digital input hannel. "n the seond ase, the system isseen as proiding enhaned memory, proessing, and ommuniation abilities to theuser. /he partiular ie2 adopted has strong impliations onerning the -lo2 o- ontrol

     bet2een the user and the system and the partiular mental operations inoled in theognitie proessing o- the user.

    ight omponents o- ognitie proessing are spei-ied in the model and are indiated by the arro2s into and 2ithin the irle sho2n in 4igure 1.3. /he -irst arro2 at the tople-t indiates the userGs attention to ertain input -rom the tas0 enironment suh asinstrutions, data, doumentation, et. /he seond arro2 moing to the right represents

     problem soling 2hih may inole planning and in-ormation proessing that ours be-ore the user inputs in-ormation to the omputer. /he third arro2 aross the toprepresents the userGs intention -or input to the omputer. "t may inole the -ormulationo- a ommand or a plan -or menu searh. /he -ourth arro2 represents the atualresponse prodution that trans-ers in-ormation to the omputer. /he user may type aommand, selet an option, or point to a sreen loation.

    /a0e a situation in 2hih the user direts his or her attention to a memo re?uesting thatHohn #mithGs telephone number be hanged to >(>'6333. /he user oneies o- thesolution to the problem as -ollo2s% 4ind Hohn #mithGs -ile, loate the telephone numberin his -ile, and then hange the number. /he intended solution must then beimplemented in terms o- the humanomputer inter-ae. /he user may -ind Hohn #mithGs-ile by seleting an option, J-ind by nameJ and then typing in the name. /o aomplishthis the user must generate the response produtions to selet this option and to type thename. @ien that the -ile is loated, the user may selet the options JupdateJ andJtelephone numberJ and -inally type the ne2 number.

    /he bottom hain o- arro2s in 4igure 1.3 starting -rom the rightmost arro2 into theirle indiates the reerse proessing o- in-ormation originating -rom the omputer and

  • 8/15/2019 computer/human interface design

    19/295

    ultimately altering the tas0 enironment. /he -irst arro2 at the right represents thedisplay o- in-ormation. /ypially, this is in-ormation displayed on the sreen but mayalso inlude displays in other modalities. /he next arro2 indiates the userGs enodingand interpretation o- that in-ormation. /he next arro2 moing le-t represents internalealuation and ognitie proessing o- the omputer output, and the last arro2 sho2s

    the -inal result o- the proess in o- supplying an ans2er to the tas0 enironment.

    At eah stage o- interation a di--erent display is generally sho2n. "n the exampleaboe, the -irst display may list a set o- -untions. /he user interprets this display byenoding the in-ormation as meaning-ul options or messages. /he option J-ind bynameJ is enoded and then ealuated as the desired -untion. 4eedba0 is also enodedand ealuated. 4or example, i- the message is JHohn #moth not -ound,J the ealuation isthat a typographi error 2as made. 4inally, 2hen the hange o- number isa0no2ledged, the user may produe an oert ans2er to the memo indiating that thetas0 2as ompleted.

    =ser harateristis a--et the proessing at eah o- these points as they depend on pereptual s0ills, attention, memory, and motor s0ills. =ser harateristis may begrouped into three types%

     Knowledge Characteristics. =sers ary in terms o- their 0no2ledge about the system."n general, 2e an no longer say that noie users hae little or no 0no2ledge o- systemoperation and that expert users do. Amount o- 0no2ledge annot be onsidered as aunidimensional attribute. "nstead, 2e must onsider more are-ully 2hat the user 0no2sabout !a* the tas0 domain in terms o- semanti and proedural 0no2ledge, !b* therepresentation o- the tas0 domain on the omputer, and !* the omputer in terms o-semanti and syntati 0no2ledge.

    Cognitive Characteristics. =sers ary in their ability to sole problems, ma0edeisions, and per-orm mental tas0s. /he assessment o- in-ormation'proessingapaities and their relationship to per-ormane has been the subFet o- muh 2or0 inognitie psyhology !e.g., unt, 19B8 #ternberg, 19BB*. An analysis o- the partiularognitie omponents inoled in a tas0 should proe use-ul in system design.

    Skill Characteristics. =sers ary in their ability to read and type text, dra2 graphiimages, point at obFets, and tra0 moing targets. /hese s0ills may be o- aryingimportane in using a system and in many ases re?uire onsiderable training and

     pratie. "t should also be mentioned that ne2 s0ills are deeloped by users throughextensie pratie.

    )*+*- Computer Processing* An analogous set o- arro2s is sho2n in the box in 4igure1.3 representing the omputer. /hese pertain to inputoutput to the enironment that isnot neessarily a part o- the humanomputer inter-ae and to internal proessing. /hearro2s that point into and out o- the intersetion bet2een the human and the omputer2ill be disussed in the next setion. "n terms o- the harateristis o- the omputer,designers must ta0e into onsideration its speed and memory and proessing apaities.o2eer, in the present ontext 2e are only interested in ho2 these harateristismani-est themseles at the humanomputer inter-ae.

  • 8/15/2019 computer/human interface design

    20/295

  • 8/15/2019 computer/human interface design

    21/295

  • 8/15/2019 computer/human interface design

    22/295

    /he 0ey to measuring subFetie ealuations is to ensure reliability and internalonsisteny o- the ratings. $y reliability 2e mean that users display little error inma0ing ratings and that i- they rated the same attributes a seond time, there 2ould be arelatiely strong relationship bet2een the t2o ratings. $y internal onsisteny 2e meanthat the ratings -ollo2 a logial relationship. 4or example, i- subFets are as0ed to

    ompare three systems, A, $, and C, and they rate A superior to $, $ superior to C, thenthey must rate A superior to C in order to be internally onsistent. ithout internalonsisteny, the meaning-ulness o- the results are in ?uestion.

    #urey data is use-ul in desribing systems, -or deteting strong and 2ea0 points, and-or suggesting improements. 4or example, in a study by Norman !196* students ratedthe oerall ease o- use o- hypothetial systems desribed by lists o- positie andnegatie attributes. /he impat o- eah attribute 2as saled relatie to all others. 4igure1.> lists the desriptions o- the attributes and graphs the impat o- eah attribute on theratings. 5ne o- the most important positie attributes, at least -or this group o- students,2as the ability o- the system to adapt to di--erent types o- users. /he most telling

    negatie attribute 2as haing a display desribed as on-using and di--iult to read.#uh ratings help to set design priorities.

  • 8/15/2019 computer/human interface design

    23/295

  • 8/15/2019 computer/human interface design

    24/295

  • 8/15/2019 computer/human interface design

    25/295

    "n the next hapter 2e ta0e an analyti loo0 at the arious types o- menu seletionsystems. A number o- -eatures and design harateristis 2ill be outlined. +any o- theseare -ators that hae been the subFet o- empirial researh and are disussed insubse?uent hapters.

    Chapter + 

    Types of Menus an% Cognitive Structures 

    A -e2 years ago the idea o- menu seletion simply meant that at some point in the program a list o- options 2ould be presented and the user 2ould selet a number orletter orresponding to the option desired. /oday, menus ta0e on a muh 2ider range o--orm and diersity o- appliation. Although the onept is muh the same, the

    appearane and the po2er o- menu seletion has hanged as exempli-ied by systemssuh as #mall/al0 !@oldberg I :obson, 19B98 /esler, 191*, 5@ !+Cra0en IA0syn 19>8 :obertson, +Cra0en, I Ne2ell, 191*, and a host o- ommeriallyaailable so-t2are that spans the 2hole gamut o- appliations and user audiene.

    "n this hapter menu struture, physial -ormat, and response re?uirements 2ill bedisussed. +enu struture re-ers to the branhing apability o- the menu. +enustruture determines the extent to 2hih the urrent seletion a--ets subse?uent options.e 2ill see that some highly onstrained menus sere a purpose in determining the

     proper se?uene o- transations bet2een the user and the system. 5n the other hand,users o-ten 2ish to or need to Fump to another atiity. +enu seletion systems in the

     past hae been ritiied beause they hae not allo2ed the user to esape -rom a partiular menu path in order to omplete some other tas0 and then return to the preious point in the menu. A -e2 systems are beginning to allo2 the user to per-ormtas0s aording to the userGs plan o- ation rather than a onstrained path o- options./hese menus allo2 the user nearly unlimited ontrol oer program -lo2 to Fump to anyloation desired.

    /he physial -ormat o- the menu may either -ailitate the user or retard per-ormanedepending on ho2 2ell it oneys in-ormation. /he -ormat should highlight the optionsand organie them in a meaning-ul 2ay to help in isual searh8 it should help to set the

    ontext o- 2here the user is in the -lo2 o- ontrol through the menu struture8 and itshould aid the user in the deision proess.

    4inally, the user must hae some means o- indiating his hoie. +enu seletionsystems ary 2idely in this respet. /he response may be a number8 a letter8 a string o-haraters8 a speial -untion 0ey8 the positioning o- a ursor using se?uene o- arro20eys, a mouse, a tra0ball, or a Foysti08 or ia a touh sreen or light pen. /he systemmust learly indiate ho2 the user is to ma0e a seletion. :esponding to menu itemsmay be dissoiated -rom the main tas0 !e.g., using a mouse to ma0e seletions in a 2ord

     proessing appliation 2here the main tas0 is 0eyboard entry* or it may be integrated2ith the tas0 !e.g., using a touh sreen to indiate moes in a hess playing program*.

  • 8/15/2019 computer/human interface design

    26/295

    "n the -ollo2ing setions arious attributes and aspets o- di--erent types o- menusystems 2ill be disussed. /he intent o- this hapter is to lay out all o- the possibilities-or designing a menu seletion system be-ore onsidering the ognitie proesses by2hih users naigate through these systems.

    +*) Menu Structures 

    /he userGs ognitie struture o- a tas0 embodies a set o- expetations about 2hathappens 2hen and 2hat leads to 2hat. e 2ill see that menu struture should beonsonant 2ith the userGs ognitie struture. /here should be a grae-ul guidane o-atiities that meshes 2ith the userGs expetations. 4or example in a -inanialtransation, i- the system as0s Jould you li0e to !1* deposit -unds or !

  • 8/15/2019 computer/human interface design

    27/295

    hen one menu -rame naturally leads to another or 2hen a set o- menus are lusteredtogether, the user begins to pereie a series o- hoies. /his series de-ines a path going-rom one menu -rame to another.

    +*)*+ Se0uential $inear Menus* Linear menus hae only one path. +enu -rames are presented in a preset order -or parameter spei-iation or -or data entry. "n essene all-lo2 o- ontrol is determined by the system. /he path o- a linear menu system mayextend -rom < to n leels and inorporates an iteration o- single menu -rames as sho2n

    in the 2hole diagram o- 4igure

  • 8/15/2019 computer/human interface design

    28/295

    the series8 it presents a single menu, and adanes to sueeding menu -rames until itreahes the terminal node. 4igure

  • 8/15/2019 computer/human interface design

    29/295

    -at, a -orm -ill'in method 2ith the restrition that items are seleted rather typed into-ields. Muestions may be s0ipped and returned to later or simply le-t blan0. 4igure

  • 8/15/2019 computer/human interface design

    30/295

    sho2n in 4igure . #ymmetri menus ary only in depth and breadth. ;epth is de-inedas the number o- leels that one must traerse to reah the terminal node. $readth isde-ined as the number o- alternaties at eah leel. /he struture o- symmetri menusan be de-ined by a series o- numbers indiating the number o- alternaties at eahleel. 4or example, the menu struture in sho2n in the top panel o- 4igure is

    indiated by 33. 5n the other hand, a > alternaties at the -irstleel, < at the next leel, and at the last leel, resulting in 6> terminal nodes. "ngeneral, symmetri menus 2ith a onstant breadth o- 0 alternaties at n leels 2ill hae0 n terminal nodes.

  • 8/15/2019 computer/human interface design

    31/295

    #ymmetri menu strutures are rare in real 2orld system o2ing to the asymmetrinature o- ategories and subategories. o2eer, they are used ?uite o-ten in empirialresearh on menu seletion beause the number o- alternaties at eah leel and thelength o- all possible paths are onstant aross the tree. hen this is the ase, it is easyto ary depth and breadth and ealuate their e--et on per-ormane. 4urthermore,

    symmetri menus hae the adantage that users an expet the same number o-alternaties at eah leel and the same number o- leels. #ometimes this onstany isimportant in routine tas0s. "t sets a simple pattern o- response on the part o- the user2ho is o-ten antiipating the next transation 2ith the system.

    Asymmetri menus, ho2eer, are the rule in most pratial appliations. /he number o-alternaties aries throughout the struture and the length o- paths to terminal nodes isnot neessarily onstant. Although menus do not hae a onstant depth and breadth,they may be desribed statistially by their aerage depth and breadth. /he amount o-ariability in the struture may be assessed by the standard deiations o- depth and

     breadth. /he de-initions o- the aerage and standard deiation -or depth and breadth,

    ho2eer, are not straight-or2ard. 5ne ould loo0 at the aerage path length -or all possible paths and the aerage number o- alternaties aross all -rames8 ho2eer, thisdoes not ta0e into aount the atual utiliation o- the menu. "- a user isits all terminalnodes 2ith e?ual -re?ueny, he experienes a di--erent aerage depth than i- he ma0esall hoies 2ith e?ual probability. Conse?uently, a 2eighting sheme needs to beapplied in assessing aerage depth and breadth. 4or aerage depth, 2e may use !a* e?ual2eighting o- all possible paths, !b* 2eighting by probability o- path gien e?ual hoie

     probability, and !* 2eighting by probability o- path gien obsered hoie probabilities. /he e?uations -or the 2eighted aerage and 2eighted standard deiation-or depth are%

    Ag. ; D OO#igma 2ini OO#igma 2i !1*

     

    #d. ; D OOradial! OO#igma 2i ! ni  ' Ag. ; *< OO#igma 2i *, !

  • 8/15/2019 computer/human interface design

    32/295

    4or breadth, one may use either !a* an e?ual 2eighting o- the number o- alternatiesaross all nodes or !b* a 2eighting by the -re?ueny o- times the node is isited. "n theseond ase, these -re?uenies may be determined by !a* e?ual path -re?ueny, !b* e?ualhoie probability, or !* obsered hoie probabilities. /he e?uations -or 2eightedaerage and 2eighted standard deiation o- breadth are%

    Ag. $ D OO#igma 2F 0F OO#igma 2F8 !3*

     

    #d. $ D OOradial ! OO#igma 2F ! 0F ' Ag. $ *< OO#igma 2F *, !>*

    2here 0F is the number o- alternaties at node F and 2F is the 2eight -or node F. /able

  • 8/15/2019 computer/human interface design

    33/295

     

    Probability eights

    Path ;epth ?ual Path ?ual Choie 5bsered Choie 

    1 1 111 1< .&&&

    < < 111 16 .1 3 111 11< .&1&&

    ( 3 111 1 .&1 111 196 .&&1(

    9 > 111 196 .&&1(

    1& > 111 196 .&&1(

    11 > 111 196 .&&1(

    Aerage 3.&9 1.B 1.

  • 8/15/2019 computer/human interface design

    34/295

    3 < 1 < 16 .& > 1 B 16 .&6&

    ( > 1 > 1 .&&6

    Aerage 3.&& 9

  • 8/15/2019 computer/human interface design

    35/295

    ylial part. Conneted graphs may also be harateried as symmetri or asymmetri.4or example, @raphs A, $, and C are symmetri and @raph ; is asymmetri.

    Conneted graphs li0e the hierarhial graphs ary in terms o- the number o- menu-rames, aerage number o- alternaties, and length o- path. stimates depend on ho2the alues are 2eighted. As be-ore they may be gien e?ual 2eights or 2eights thatdepend on menu use and probability o- paths. A maFor di--erene bet2een onnetgraphs and hierarhial menus is that in onneted graphs there may be multiple paths

     bet2een t2o nodes. +enus ary in terms o- onnetiity, that is, extent to 2hih nodes

    are lin0ed by multiple paths.

  • 8/15/2019 computer/human interface design

    36/295

    /he onneted graph is generally used to proide the user 2ith a -ull sense o- ontroloer the -lo2 o- ations. /hese systems are -ound most o-ten as menu replaements o-ommand languages -or operating systems and appliation programs. 4or the ontrol o-

     program -lo2, the onneted graph is typially se?uential. /he options open to the userare limited to the number o- paths leading -rom the urrent node. hile this is limitation

    is help-ul in proiding -ous -or the user, it may also proe -rustrating 2hen there is aneed to per-orm a onurrent tas0 unrelated to the urrent loation in the graph or a needto brea0 the urrent -lo2 o- ontrol and Fump out o- the graph to selet some ne2 entry

     point. ent trapping menus allo2 -or this leel o- ontrol are are desribed in the nextsetion.

    +*)*2 'vent Trapping Menus* ent trapping menus proide in essene a set o-simultaneous menus superimposed on a se?uential hierarhial or onneted graphmenu system and may be atiated at any point in a tas0. /ypially, these menus appearas speial -untion 0eys or as pull'do2n or pop'up lists o- options. "n the ase o- speial-untion 0eys, the user selets a menu or item in a menu by pressing a speially denoted

    0ey. "n the ase o- pull'do2n or pop'up menus a ursor is plaed on the menu label andthe items are reealed. 4igure

  • 8/15/2019 computer/human interface design

    37/295

    !* 4inally, it may exit the urrent enironment and moe to a totally ne2 enironment.4or example, the user may selet J?uitJ to return to the operating system or seletamong other appliation programs. ent trapping menu systems ary in term o- thenumber and mix o- these three types o- items. /he greater the -lexibility and omplexityo- a system, the greater the number o- eent trapping options.

    ent trapping menus proide an eer present ba0ground o- ontrol oer the systemstate and parameters 2hile the user is 2or0ing on a -oreground tas0. /he menus areonstantly aailable and help to establish a sense o- ontext 2hile things are hanging inthe -oreground. /he menu bar may hange -rom one appliation program to another, buttypially there is some degree o- ommunality bet2een the eent trapping menus andsome degree o- distintieness that helps to remind the user 2here he is. o2eer, i- themeaning o- speial -untion 0eys and the partiular items in pull'do2n menus arygreatly or hange 2ithout 2arning, the user may experiene loss o- position ordisorientation and ma0e habitual errors. e 2ill see in the next setion ho2 the usermust be 0ept in-ormed o- the urrent state o- the system in terms o- in-ormation

    oneyed to the user by the menu -rame.

    +*+ Menu 3rames 

    /he menu -rame onsists o- the ontext, the stem, the leaes, and the responseinstrutions. Needless to say, menu seletion systems ary drastially in terms o- 2hatin-ormation is presented, the amount o- in-ormation, and the -ormat o- the -rame. /hedesign o- a partiular system must ta0e into onsideration ho2 muh in-ormation theuser needs to 2or0 meaning-ully 2ith the system. "n general, one 2ould expet thatusers 2ho are un-amiliar 2ith the menu need more in-ormation. As users beome moreand more experiened, erbose desriptions may be replaed 2ith short mnemonireminders. "n the -ollo2ing setions ariables pertaining to the amount and type o-in-ormation in the menu 2ill be disussed.

    +*+*) Contet Information* /he ontext proides -eedba0 that tells the user 2here heis in the proess, 2hat the past hoies and outomes 2ere, and possibly ho2 muh-urther it is to the terminal node. Context in-ormation is extremely important in omplexmenus 2here the user may experiene a subFetie loss o- position and assoiateddisorientation. /he user may not reall ho2 -ar do2n the tree he or she has ome, orho2 suess-ul past hoies hae been in getting loser to the goal, or ho2 many morehoies must be made be-ore reahing to the goal. "t is too o-ten the ase that hal-2ay

    through a searh, the user as0s, JNo2 2hat 2as it that " 2as loo0ing -orEJ umanmemory, being 2hat it is, needs assistane.

    At a minimum, the -rame should proide in-ormation that the user is aessing the right-untion. "n a ban0 transation, i- the ustomer has seleted his he0ing aount, thenall -urther hoies !balane, 2ithdra2, deposit* should be learly in the ontext o- thehe0ing aount. /his -eedba0 proides the user onstant assurane that he is on theright tra0. 5n the other side, ontext in-ormation should not be oerly erbose. "- it is,it obsures the urrent hoie and may add to the on-usion.

    Context may be established by arious methods o- lin0age bet2een suessie menu

    -rames as illustrated in 4igure

  • 8/15/2019 computer/human interface design

    38/295

    /emporal se?uene implies a ause'e--et relationship8 ho2eer, it does not indiate thenature o- the relationship bet2een one -rame and the next. /he ontext established bytemporal lin0age deays 2ith time and may need to be maintained by other methods o-lin0age.

    erbal lin0age may be aomplished by using the preiously seleted option as the title-or the next menu -rame. /he title may be umulatie as one goes -urther along a path.4or example, in a ideo text serie the title o- a menu -rame may be JNe2s% Ne2 7or0/imes% orld Ne2s% "ndia% /odayJ indiating the path to the urrent set o- options2hih may be a list o- stories.

    #patial lin0age may be aomplished by graphi methods that presere temporal lin0ageand enhane erbal lin0age. "n the bottom panel o- 4igure

  • 8/15/2019 computer/human interface design

    39/295

    the stem might be J#elet baud rate%J /he stem may be merely the urrent hoie or itmay inlude a reord o- all hoies in the path up to the present hoie. "n nestedhierarhial menus, the stem in-ormation may inrease 2ith eah hoie. #eletionsmay inlude a trae o- the 2hole path do2n to the urrent point. 4or example, along thelines o- the ;e2ey ;eimal #ystem a stem in an in-ormation retrieal system might be%

    istoryQ=.#.QCiil arQonomiQ"mpat onEQ.

    o2eer, in generating stem in-ormation, there is a trade'o-- bet2een spei-iity andlutter. hen one ould easily get lost in the mae o- hoies, umulatie stemin-ormation is essential. "n other ases, ho2eer, the user may not need to see all o- thestem in-ormation. 5nly that 2hih is li0ely to be lost -rom short term memory and thatis ruial to spei-ying the urrent hoie must be displayed.

    #tem in-ormation should proide the noie user 2ith a rationale -or 2hy this hoie is being made and 2hat itGs impat 2ill be on subse?uent proesses. 4or the set o- hoiesJ@othi, :oman, +odern,J a stem suh as J#elet one o- the -ollo2ing%J is deoid o-

    in-ormation. /he noie user does not 0no2 2hat they re-er to or 2hat the hoie 2illa--et. 5n the other hand, a stem suh as J#et the type -ont to one o- -ollo2ing -or

     printing doument%J proides that in-ormation. :esearh in ognitie psyhology sho2san enormous impat o- proiding a meaning-ul title on a personGs interpretation andunderstanding o- subse?uent in-ormation !$rans-ord I Hohnson, 19B

  • 8/15/2019 computer/human interface design

    40/295

    +*+*. (esponse Information* 4inally, response instrutions tell the user ho2 to indiatethe hoie. 4or example, it may say Jnter the number o- your seletion and press thereturn 0ey,J or simply JPress 0eyJ 2hen speial -untion 0eys are labeled. xpliitinstrutions are o-ten needed -or -irst time users o- the system. ith experiene, usersgain a sense o- 2hat is expeted o- them by the system. 5erly erbose instrutions

    generally on-use the user or are simply not read. "nstrutions -or response proedures2hih are di--erent -rom the usual mode o- interation need to highlight thosedi--erenes. 4or example, a response may be reognied and proessed by the system!a* immediately a-ter a single 0ey is pressed, !b* only 2hen terminated by pressing thereturn 0ey, or !* only 2hen terminated by pressing a speial JsendJ 0ey. "- usershabitually terminate lines 2ith a return 0ey, errors and on-usion may our i- adi--erent type o- response is re?uired.

    :esponse in-ormation may also inlude -eedba0 or eri-iation that a partiularalternatie 2as atiated or is about to be atiated. #ystems ary -rom proiding no-eedba0, to a single auditory beep that some response 2as reorded, to on-irmation o-

    the response in an input -ield or by highlighting the seleted alternatie.

    +*+*1 Information 3ormat* "n early alphanumeri appliations o- menus, in-ormation2as presented se?uentially in the -ollo2ing order% ontext, stem, lea-, response. /hethree panels o- 4igure

  • 8/15/2019 computer/human interface design

    41/295

    ith the adent o- graphis and sreen based systems, other methods o- displayingin-ormation hae been used. Context is o-ten gien by the ba0ground on 2hih themenu appears. A pop'up 2indo2 or pull'do2n menu appears oer the 2or0 at hand8

    thereby retaining a sense o- position. /he stem and lea- in-ormation in these systemsremains muh the same exept that options may be graphially displayed using ions.5-ten the stem in-ormation may be impliit rather than expressly stated. "t is expetedthat the layout o- suh in-ormation 2ill greatly a--et the userGs ability to omprehendthe -untioning o- the system.

    +*- (esponse Mo%e 

    /he method o- response seletion is an important one and aries substantiely amongsystems. e 2ill see later that there are a number o- ognitie impliations regardingthe response mode. #ome response modes re?uire a mapping o- the alternatie to a ode

    and the manual entry o- that ode. 5ther systems re?uire a manual pointing to thealternatie. #ome allo2 -or eri-iation o- the response be-ore atiating it and others

  • 8/15/2019 computer/human interface design

    42/295

    donGt. /he setions that -ollo2 disuss these and other ariables as they pertain to themaFor -orms o- response input.

    +*-*) Stan%ar% 4eyboar% Input* Keyboard input re?uires the user to identi-y thealternatie he 2ishes to selet and then determine the appropriate 0eystro0es to ma0e

    that seletion. +enus may use !a* se?uentially numbered alternaties, !b*nonse?uentially numbered alternaties, !* alphabetially labeled alternaties, !d*randomly labeled alternaties, or !e* mnemonially labeled alternaties.

    /he 0eyboard response may or may not re?uire a terminating harater suh as thereturn 0ey or send 0ey. "- it does, it proides the user 2ith the opportunity to eri-y thathe has entered the orret label. "- it does not, the response re?uires one less 0eystro0e./he trade'o-- is bet2een speed o- operation and ost o- errors. "- errors an easily and?ui0ly be orreted 2ithout on-using the user and i- speed is o- the essene, then sel-terminating responses are alled -or. 4or noie users, response eri-iation isimportant8 ho2eer, on-usion and -rustration an arise 2hen the user does not 0no2

    that he must press the return 0ey to ontinue.

    #tandard 0eyboard input re?uires some -amiliarity 2ith the layout o- the 0eys.Per-ormane is -ailitated, i- the user 0no2s ho2 to type and does not hae to searh -or the 0eys. Number 0eys hae the adantage o- being se?uentially plaed along the topro2 o- 0eys or on a separate numeri 0eypad. +ultiple harater abbreiations are themost di--iult -or the nontypist.

    +*-*+ Special 3unction 4eys* #peial -untion 0eys are used in many systems insteado- standard 0eyboard input. /he 0eys may only be labeled se?uentially !e.g., J41J, J4

  • 8/15/2019 computer/human interface design

    43/295

    -ouses on the sreen as a diret input deie, the other -ouses on the 0eypad as anindiator o- options.

    hen the options are listed on the sreen, it seems natural to be able to point at or touhthe desired item. /he sreen beomes in essene the 0eypad. ;iret seletion may be

    aomplished using a light pen or a touh sreen. /he use o- a light pen re?uires that theuser pi0 up the pen and diret its sensor at the desired item and press a selet button./he touh sreen is less enumbering and merely re?uires that the user physially touhthe item on the sreen. /he adantage o- a touh sreen is the simpliity o- seletion.Among the disadantages is the problem that the -inger an obsure the item, theauray o- pointing may be lo2, and the user may experiene arm -atigue a-ter long

     periods o- interation.

    "ndiret pointing methods may be used to selet items on irtual 0eypads as 2ell."ndiret pointing is done by the moement o- a ursor aross the items in the menu. /heursor may be direted by a ontinuous analog input deie suh as a mouse, a

    tra0ball, or a Foysti0. Alternatiely, the ursor may be plaed by disrete presses o-arro2 0eys or a spae bar. ith ontinuous moement o- the ursor, the user must beable to selet the target item by plaing the ursor into a spatial -ield and pressing aselet button. @enerally, isual -eedba0 is re?uired suh that the item turns reerseideo 2hen the ursor is on that item. hen the ursor is not on an item and the selet

     button is pressed, generally no ation is ta0en. /he adantage o- ontinuous inputdeies is that users an rapidly moe the ursor to the target in a ontinuous moe. /hedisadantage is that good eye'hand oordination and hand steadiness is re?uired and theuser may miss the target. /he user must grab the input deie, -ind the urrent loationo- the ursor, loate the target, plot a traFetory -rom the Point A to Point $, eri-y 2henthe ursor is on target, and then press the selet button.

    ith disrete input deies, a number o- 0eystro0es may be re?uired to moe the ursor to the desired item. /he adantage, ho2eer, is that most systems Fump the ursor -romitem to item so that the ursor is al2ays on a target and neer on an unde-ined hoie.Arro2 0eys an be used to moe up or do2n a ertial list or moe to the le-t or right ina horiontal list. Arro2 0ey seletion re?uires the user to loate the arro2 0eys, -ind theurrent the position o- the ursor !or highlighted alternatie*, loate the target, pressdiretion 0eys until the ursor is on target, and then press the selet button.

    A disadantage in many ases in using the sreen as the input deie is that the menu

    options obsure or disrupt the proess o- 2or0. A irtual 0eypad that automatiallylabels 0eys depending on the urrent set o- options, soles this problem. /he 0eypad inessene beomes an output deie listing the urrent menu -rame. /he atiealternaties are indiated by LC; or L; displays on or next to the 0eys. /his solutionis analogous to haing a separate 2indo2 or seond sreen -or menu seletion.

    +*. Summary 

    A surey o- menu systems indiates that they di--er along a number o- -ators andharateristis. "n this hapter menus 2ere harateried by !a* the struture o- the menutree, !b* type and amount o- in-ormation presented in the menu -rame, and !* the mode

    o- response seletion. /he many -ators that go into designing a menu system de-ine aspae o- all possible systems. /he ideal system must blend the right leels o- -ators and

  • 8/15/2019 computer/human interface design

    44/295

  • 8/15/2019 computer/human interface design

    45/295

    o- menu items that eo0e ations or hanges. An analysis o- the tas0 is neessary -or thedesigner to determine 2hat -untions need to be implemented. "n deeloping ataxonomy, one may ta0e either a Jtop'do2nJ or a Jbottom'upJ approah !Chin, 196*."n a top'do2n approah, the designer lists maFor top'leel -untions 2hih are thenre-ined in greater and greater detail. "n a bottom'up approah, the 2hole list o- spei-i

    -untions is analyed and organied into groups. hih approah is used depends on the partiular tas0 and -untions. 4or example, in text editing one may start 2ith a top'do2napproah as sho2n in 4igure 3.1. /he maFor tas0 omponents are !a* -ile -untions, !b*

     printing, !* text modi-iation, !d* -ormatting, and !e* bro2sing. ah o- these may then be re-ined to more spei-i -untions. /he top'do2n approah has the adantage that as-untions are added they are inorporated in a hierarhial struture.

    5n the other hand, 2hen a list o- spei-i -untions already exists, the bottom'upapproah may be used. /as0s that hae been manually per-ormed in the past may beanalyed into subtas0s. /he subtas0s are then lustered into groups. 4igure 3.< gies anillustration o- a tas0 analysis o- a message handling system.

  • 8/15/2019 computer/human interface design

    46/295

  • 8/15/2019 computer/human interface design

    47/295

  • 8/15/2019 computer/human interface design

    48/295

  • 8/15/2019 computer/human interface design

    49/295

  • 8/15/2019 computer/human interface design

    50/295

    The S!stem State Transition 'iagram Should (ptimi)e the User Subtask Transition

     $atrix& A state transition diagram spei-ies a set o- states !indiated by irles*, the possible transitions -rom one state to another !indiated by arro2s*, and the onditions

    under 2hih a transition ours. "n menu seletion systems, the allo2able transitions aredetermined by the menu struture and the onditions -or transitions are determined byuser seletion. /o per-orm a tas0, users all upon a number o- subtas0s. @ien the set o-subtas0s, a transition matrix spei-ies the probability that one subtas0 -ollo2s another.

    /he upper panel o- 4igure 3.> gies an example o- a probability transition matrix -or seto- subtas0s -or playing a omputer game. 4or eah subtas0 implemented at #tate n!sho2n in the ro2s*, there is a set o- probabilities that the user needs to go to partiularsubtas0s at #tate nR1 !sho2n in the olumns*. 4or example, i- the user is at #ubtas0 A,there is a .( probability o- going to #ubtas0 $ and a .( probability o- going to #ubtas0C. @ien this probability matrix, one may ealuate the graph struture o- a menu system2hih implements these states. A good menu should allo2 diret paths -or high

     probability transitions. 5n the other hand, lo2 probability transitions need not be diret, but may re?uire suessie seletions.

  • 8/15/2019 computer/human interface design

    51/295

  • 8/15/2019 computer/human interface design

    52/295

    unneessary paths up the tree are eliminated and the high probability lateral transitionsare inorporated in the struture.

    /he degree to 2hih a menu struture is improished an be estimated -rom the probability transition matrix. 4or a gien number o- subtas0s that hae to be per-ormed,

    one an estimate the perent o- the expeted number o- transitions re?uired by animpoerished menu relatie to a menu that implements a path -or all non'erotransitions. 4or example, the perent o- extra traersals -or the Jinade?uateJ menu in4igure 3.> is 6&T oer that o- the Jade?uateJ menu.

    5n the other hand, menus may be oerly rih 2hen they inlude a number o-super-luous or underutilied paths in the struture. An oerly rih menu is sho2n at the

     bottom o- right 4igure 3.> 2hih allo2s a transition -rom any node to any other node.Although a rih menu minimies the number o- -rames traersed, the inreased numbero- alternaties may retard hoie and response times o- the user. "mpoerished menus2ill eidene -aster response times per -rame, but the oerall time to per-orm a tas0 may

     be longer sine more responses are re?uired. ;esigners must trade'o-- the osts bet2eenrih and lean menus to optimie per-ormane.

    /he e--iieny o- menus an be ompared by estimating expeted user response timesaross the menu. :esponse times are a -untion o- the number o- alternaties in eah-rame. 4or the present, assume that response time is gien by a po2er la2% : D n p R 182here n is the number o- items per -rame and let p D .B. /able 3.1 gies the responsetimes at eah node o- the three menu graphs sho2n in 4igure 3.>. =sing the transition

     probabilities also in 4igure 3.> and the graphs, the expeted response times 2erealulated and are sho2n at the bottom o- /able 3.1. /he expeted response time -or theJinade?uateJ menu is muh longer due to the -at that 6&T more responses are re?uired-or the desired state transition. /he Joerly rihJ menu also sho2s a higher responsetime due to slo2er responses at eah -rame. "t an be seen that hanges in the menustruture an result in rather large hanges in per-ormane. #imilarly, designers shouldompare alternatie implementations o- menus or iteratiely ary menu strutures to-ind one that optimies per-ormane.

    /able 3.1

    xpeted =ser :esponse /imes -or /hree +enu /ypes

    "nade?uate Ade?uate 5erly :ih

     

    A

  • 8/15/2019 computer/human interface design

    53/295

    4 . >.&9

    -*. 3unctions of Menu Selection 

    "n this setion the spei-i -untions o- a menu seletion 2ill be disussed. ahseletion made by the user in a menu per-orms one or more -untions. /he -untion may

     be merely to branh to another menu or it may be to do something else. /his setion 2illde-ine the -our -untions o- !a* pointing, !b* ommand ontrol, !* output, and !d* input.5ther taxonomies o- menu -untion ould be spei-ied based on program operation8ho2eer, the present one 2ill proe use-ul -rom the perspetie o- understanding

    ognitie ontrol by the user.

    -*.*) Pointing: Moving to a e6 o%e* ah seletion by the user branhes to asuessie node in the menu tree. Although this is an inherent -untion o- all menus, theimportane o- this -untion aries among systems. "n some ases, it is the sole purposeo- the seletion to traerse the menu tree. #eletions sere in a step2ise spei-iation o-an obFet or ommand. /he oerriding purpose is one o- pointing to something. 4romthe perspetie o- the user, menu seletion is li0e naigation. /he sole purpose o- eahseletion is to steer the system to2ard a destination. "n this 2ay the proess is goaloriented. Although seletion errors ta0e the user o-- tra0 and lead to a loss o- time, theyare not destrutie and are in general undoable.

    /he pointing -untion is essential in hierarhial menu systems. An example 2ould be amenu system -or a timesharing utility. A number o- series are aailable to the user8 butin order to aess one, the user must spei-y its name. :ather than haing to type in thename o- the serie, the user is gien a series o- hoies. /he series are typiallylustered by type into a hierarhy. ah seletion by the user does nothing exept moeto the next leel o- spei-iation. 5ne the user omes to a partiular serie on thesystem, the -untion o- the menu hanges -rom traersal to implementation.

    -*.*+ Comman% Control: 'ecuting a Proce%ure* /he seond -untion o- seletion is

    to diret the omputer to exeute a proedure or to implement some ation. "n this ase,the hoie eo0es a proedure that per-orms some -untion. "t may open or lose a -ile,2rite or read to a -ile, trans-orm data, et. As noted, this may only our a-ter the userhas traersed the menu tree and is at some terminal node. 5n the other hand, eeryseletion made by users may initiate a program proedure as they traerse a menunet2or0. "n this ase, suessie seletions eo0e a se?uene o- proedures. /he usersdetermine the seletion and order o- proedures. 4rom the userGs perspetie, seletionsare Jgo'aheadJ ommands -or the system to exeute a -untion. "t should be lear tousers that seletions are not merely pointing to items, but that they are atiatinghanges. rrors in ommand ontrol may not be so inonse?uential as pointing to the2rong node. Conse?uently, the user needs to be in-ormed i- seletions are undoable

     be-ore they are seletedd.

  • 8/15/2019 computer/human interface design

    54/295

  • 8/15/2019 computer/human interface design

    55/295

    aing said this, it is also true that systems may emphasie only one -untion. +enus-or time sharing systems, -ile management, and doument retrieal systems lean moreto2ard traersal o- a data base ia the pointing -untion. +enus -or operating systemsand appliation programs suh as 2ord proessors and spread sheets lean more to2ardexeuting proedures ia the ommand ontrol -untion. +enus -or help systems,

    tutorials, and in-ormation retrieal systems lean more to2ard displaying in-ormation iathe display -untion. 4inally, menus -or online ?uestionnaires and data entry systemslean more to2ard data and parameter spei-iation ia the input -untion.

    /he 2ay in 2hih a menu is designed should be ommensurate 2ith the pereied tas0and the 2ay in 2hih the system is used to aomplish the tas0. +a@regor, Lee, andLam !196* distinguish bet2een JommandJ menus and JideotextJ menus and notethat ommand menus are designed to selet a relatiely limited set o- items !3& to 1&&*.xperiened users hae learned the positions o- menu items. 5n the other hand,ideotex menus are li0ely to aess relatiely large data bases 2ith thousands o-douments re?uiring a large number o- menu -rames. "n ommand menus the user

    generally does not read all o- the items but merely sans to the desired one and seletsit. "n ideotext menus the user may need to read eah item are-ully in order todetermine the next seletion. hile ommand menus may be designed 2ith a relatielylarge number o- options, ideotext menus aording to Paap and :os0e'o-strand!196* should -ous the user do2n a narro2er path by presenting a smaller number o-items distributed aross more menu -rames.

    +enus that merely sere to point suessiely to other other -rames should be designedto -ailitate traersal possibly by Fump ahead ommands and they should proide theuser 2ith a sense o- position possibly by displaying a path or a global map o- the menutree.

    -*1 "peration by Menu Selection: Comman% Menus 

    +any appliations use menu seletion as a replaement -or ommand languages. :ather than spei-ying a ommand by entering it ia the 0eyboard, the ommand is seletedalong 2ith its operand ia menu seletion. Command menus hae the adantage o-oeroming t2o inherent problems o- ommand languages. 4irst, menus help todisambiguate terms that an hae multiple meanings. A term suh as list  may be aommand to display a list o- items or it may be the name o- an obFet. /he partiularmeaning o- a term an be determined by the ontext. /he seond problem is that a

    -untion ould hae a number o- e?ually li0ely names. +enus resole 2hih o- anumber o- synonyms is reognied by the system as a ommand. 4or example, the-untion o- terminating a proess may be labeled% stop, halt, ?uit, end, -in, bye, 0ill,term, et. ith a ommand language, one must remember 2hih term is appropriate.

     Natural language systems may reognie a number o- synonyms8 ho2eer, users muststill go through the ognitie proess o- generating a term and 2ondering i- it 2asorretly reognied. 4urther di--iulty arises i- t2o or more synonyms result indi--erent ations by the system. 4or example, J?uitJ may exit a program and return tothe operating system, JhaltJ may stop the proessor until a ontinue ommand is issued,J0illJ may terminate a proess 2ithout storing the results, and JendJ may grae-ullyterminate a proess and store the results. $y listing all -our ommands as menu items

    the user is alerted to the -at that eah has a di--erent e--et.

  • 8/15/2019 computer/human interface design

    56/295

    +enus -or ontrol o- operations usually inole a se?uene o- seletions -rom !a* a seto- ommands, !b* a set o- operands, and !* a set o- options or parameter alues.Commands are essentially synonymous 2ith erbs, operands 2ith names o- obFets, andoptions 2ith ?uali-iers suh as adFeties and aderbs. /he se?uene o- seletions to

     per-orm a tas0 onstitutes a sentene 2hih must on-orm to the grammatial order

    re?uired by the system. 4or example, the ommand to open a spei-i -ile ould beimplemented by -irst seleting the option J5penJ -rom a list o- operations and thenseleting a -ile -rom a list o- names o- -iles. ;epending on the appliation, a di--erentorder o- seletions -or the ommand, the operands, and the options may be re?uired. Adoument !operand* may be seleted -irst, then the ommand !print*, and -inally, the

     print options !e.g., dra-t, letter ?uality*. 5n the other hand, the appliation may re?uirethat the print options and the doument be spei-ied prior to the print ommand. 4orommand languages researh suggests that the ommand preede the operand !$arnard,ammond, +aLean, I +orton, 19

  • 8/15/2019 computer/human interface design

    57/295

  • 8/15/2019 computer/human interface design

    58/295

    items in the list is that the presene and loation o- items is -ixed. =sers are exposed toall the items and are in-ormed as to 2hih are urrent. hat they may not 0no2,ho2eer, is ho2 to s2ith modes so as to atiate desired items.

    "tems may also be s2ithed or rephrased depending on the urrent state. "n one state themenu may display the item J5pen 4ile.J "n another state, it may hange the item toJClose 4ile.J 5r it may display both J5pen 4ileJ and JClose 4ileJ 2ith one or the othergrayed out. /he -irst -orm allo2s -or shorter lists, but may on-use een the experiened

    user 2ho must are-ully he0 the urrent state o- the menu be-ore seleting items.hen menus s2ith options depending on ontext, it an be -rustrating to the user 2ho

  • 8/15/2019 computer/human interface design

    59/295

  • 8/15/2019 computer/human interface design

    60/295

    -*2 Decision by Menu Selection: Decision Menus 

    +enu systems an e--etiely be used to assist users in ma0ing deisions. ;eisionma0ing is a omplex tas0 2hen the number o- ompeting alternaties is -airly largeandor 2hen alternaties are omposed o- many positie and negatie onse?uenes.hen the deision an be strutured as a series o- se?uential hoies, menu seletion

     proides an appealing method o- simpli-ying the tas0. #uh systems present a series o-hoies that help -ormulate the deision. /he deision may be as simple as 2hih help-ile to aess or as omplex and ritial as deiding among alternatie ourses o- ationto resue a stranded ehile in outer spae.

    ;eision systems may be implemented using menu seletion in seeral di--erent 2aysdepending on the -ormal properties o- the deision and the method o- handling deisiontrees and matries o- alternaties.

    -*2*) Decision Trees* 5ne 2ay o- handling omplex deisions is to use deisionanalysis to deompose a omplex deision into a hierarhial deision tree. /he deisiontree represents eah se?uential hoie as a node in a graph. Alternatie ations on part o- the human are represented as ars pointing to other hoie nodes or to states o- the2orld. Alternatie onse?uenes in the enironment are also represented by ars leadingto subse?uent hoie nodes or states o- the 2orld. "n a sense, deision ma0ing thenentails traersing the deision tree by ma0ing a seletion at eah hoie point. /headantage o- implementing the deision on a menu system is that hoies are laid out ateah leel and the user an easily traerse the deision tree by ma0ing seletions.o2eer, the shortoming is that the deision ma0er may -ous on only one hoie

    node at a time and ignore more remote onse?uenes. "n general, the goal o- deisionanalysis is to proide a more omprehensie analysis in order to loate an optimal

  • 8/15/2019 computer/human interface design

    61/295

    terminal node. /o some extent menu systems may aoid this problem by displaying oneor seeral plies do2n the deision tree. o2eer, 2hen extensie searh through thedeision tree is re?uired, other approahes are alled -or suh as $ayesian analysis ortree searhing tehni?ues.

    hen menu seletion attempts to implement a deision tree, it is should be ealuated onho2 2ell it aptures the deision proess and 2hether it has hidden onse?uenes. "t iseasy to misrepresent the deision spae in the menu tree and lead the user do2n aJgarden path.J 4igure 3. gies an example that is not atypial o- many suhappliations. /he extremely ompliated deision o- areer hoie is laid out in terms o- interests and abilities. #uh systems may be prooatie and een instrutie in the 2aythey eluidate a number o- deision -ators. $ut they are o-ten misleading in that earlyhoies hae unantiipated onse?uenes on subse?uent options. Although users may beambialent about early hoies, they are neertheless -ored to ommit themseles toone path, thereby eliminating other paths that 2ould hae led to more desirable options.#uh systems hae the potential o- atually degrading deision ma0ing ability.

    o2eer, properly used, menu seletion may apture the deision se?uene in a 2aythat expliates a omplex deision proess in a straight-or2ard and omplete manner to-ailitate the deision proess. "n general, one 2ould expet menu seletion to be mostappliable 2hen the struture o- the deision tree is 2ell 0no2n or learly determined

     by a presribed rule !#imon, 19(6*.

  • 8/15/2019 computer/human interface design

    62/295

    -*2*+ Decision Matrices. +any deisions inole a seletion among a omplex set o-alternaties. ;eision theory proides a -ormal method o- representing alternaties assets o- attributes. Alternaties an be thought o- as multidimensional etors o-attributes. Consider running shoes, -or example. /hey may be desribed by the

    attributes o- prie, 2eight, midsole onstrution, outsole onstrution, type o- last, andheel li-t. "n turn, physial attributes result in subFetie pre-erenes along the attributessuh as oerall om-ort, sho0 absorption, -lexibility, toe room, oerall stability, heelounter, s2eat loss, et. /able 3.> -or example, lists 1( running shoes 2ith their prieand Fudges ratings on seen attributes. /he deision tas0 is di--iult een in this simplease o- deiding on running shoes. ;eision theory in onFuntion 2ith menu seletion,ho2eer, an simpli-y the tas0 onsiderably.

    /able 3.3

    5bFet'Attribute'=tility +atrix -or :unning #hoes

    !-rom Consumer :eports, 5tober 196, p 6(>'6((*

  • 8/15/2019 computer/human interface design

    63/295

       

    $rand and +odel Prie A $ C ; 4 @

       

     Ne2 $alane >B& U(3 ( > ( ( 3 3 <

    Asis /iger pirus U& ( ( ( > > > (

    /urnte Muantum Plus U6& > > ( ( > > >

    #auony #hado2 U(9 ( ( ( > < > (

    Asis /iger =ltra 1&&& U(B ( > ( > 3 > (

    #auony Ameria U> > > > > 3 > (

    $roo0s /rilogy UB& ( > > > 3 3 3

     Ne2 $alane 13&& U13& > ( > > > > 3

    Adidas )(&& UB( > > 3 3 ( > <

    $roo0s Chariot U6< 3 3 3 > ( ( 3

    :eebo0 ;L(6&& U(6 > > ( > 3 3 >

     Ni0e enue U6& 3 3 3 > > > >

     Ni0e ortex U6& 3 3 > 3 3 > 3

     Ne2 $alane (B( U66 3 > 3 > > > >

    toni +irage U>( 3 3 > > < < >

       

     Note. ( D better, 1 D 2orse, A D oerall om-ort, $ D sho0 absorption, C D -lexibility, ;D toe room, D oerall stability, 4 D heel ounter, @ D s2eat loss.

    ;eision theory operates on the assumption that a ?uanti-iable utility measure exists -oreah attribute. ah Alternatie i an then be desribed by an array o- numbers%

    Ai D Vui1, ui

  • 8/15/2019 computer/human interface design

    64/295

    ;eision ma0ing then inoles a proess by 2hih attribute utilities are ompared andombined in order to determine the pre-erred alternatie. ;eision theory hasinestigated a number o- possible methods -or doing this !+ontgomery I #2enson,191*. A -e2 o- these are outlined belo2. ah re?uires a di--erent leel o- assessmento- pre-erene and ma0es di--erent assumptions onerning ho2 to integrate the releant

    attributes.

    (rdinal 'ominance Rule. An alternatie is seleted i- -or eery releant attribute thedeision ma0er pre-ers that alternatie or at least does not pre-er another alternatieoer that one. /o implement the pre-erene rule in menu seletion, the system 2ould

     present a menu -or eah attribute listing the aailable leels. ;eision ma0ers 2ouldselet the pre-erred leel -or eah and the system 2ould identi-y the alternatie!s*

     possessing those attributes. =n-ortunately, it is usually the ase that suh idealombinations do not exist and no alternatie ompletely dominates the rest. "n /able 3.>-or example, there are no running shoes that are the least expensie and hae the highestratings on all the attributes. Conse?uently, one must in some 2ay relax the

    re?uirements.

     $inimum Criteria Rule. An alternatie is seleted i- it meets or surpasses a minimumriteria on eah attribute set by the the deision ma0er. /his rule 2ould again beimplemented by presenting menus to the deision ma0er -or eah attribute listing theaailable leels. $ut instead o- seleting only the pre-erred leel, the deision ma0erselet all aeptable leels !e.g., prie less than U6&, oerall om-ort S 3, et.*. "- noalternaties are loated, the deision ma0er may iteratiely relax the riteria until analternatie is seleted. "- a number o- alternaties meet the riteria, the deision ma0ermay raise the riteria.

    /he minimum riteria rule is similar to the Jsatis-iingJ rule disussed by #imon !19&*in 2hih the deision ma0er selets the -irst alternatie that meets a set o- minimumriteria. /he di--erene is that the satis-iing rule is sel-'terminating8 2hereas, theminimum riteria rule re?uires that the system per-orm an exhaustie searh through theentire set o- alternaties.

    "t is usually the ase that some attributes are more important than others. Neither o-these t2o rules ta0es this into onsideration. An alternatie may be dropped beause itdoes not meet a riterion on a releantly unimportant attribute. /he remainder o- therules belo2 ta0e into onsideration attribute importane.

     Lexicographic Rule. An alternatie is seleted i- it tends to be pre-erred oer otheralternaties on attributes Fudged to be o- greater importane than others. /o implementthis rule the system 2ould -irst present a menu o- all o- the attributes and the deisionma0er 2ould selet the most important. /hen the leels on that attribute 2ould be

     presented and deision ma0er 2ould selet the pre-erred leel. /he system 2ould thenlist the remaining attributes and the deision ma0er 2ould selet the next mostimportant attribute. /he leels on that attribute 2ould be listed and the deision ma0er2ould selet the pre-erred leel, and so on. 4or example in /able 3.>, oerall om-ortmight be seleted -irst, and rating o- ( pre-erred. Prie might be next in importane, andamong those haing a rating o- ( on oerall om-ort, the prie o- U(B is pre-erred. /he

    deision proess 2ould stop there sine there is only one pair at that prie.

  • 8/15/2019 computer/human interface design

    65/295

     *limination b! spects. An alternatie is seleted i- it remains a-ter the deision ma0erhas se?uentially eliminated alternaties not possessing desired attributes. A series o-studies by Kahneman and /ers0i !191* suggest that in many situations people ma0edeisions by through a proess o- elimination. /he deision ma0er onsiders oneattribute at a time. Alternaties that do not possess a desired aspet are dropped -rom

    onsideration. /hose that possess the aspet are retained. /he proess ontinues until all but one alternatie remains. /he elimination by aspets rule orresponds losely to the proess o- menu seletion. /he rule may be implemented by in a 2ay similar to thelexiographi rule. /he di--erene is that the rule is to eliminate alternaties not

     possessing an aspet as opposed to retaining alternaties haing pre-erred leels o-attributes.

    /here are seeral disadantages 2ith the elimination'by'aspets rule. #trit adherene tout'o--s an lead to less than desirable outomes. 4or example, the -irst attribute may be

     prie and the deision ma0er selets the aspet Jless than U6&.J "n this ase /urnteMuantum Plus is eliminated although it is only one dollar more than #auony #hado2. A

    seond problem ours 2ith the desirability o- one attribute is ontingent on the leel o- another. 5nly through ba0tra0ing and extensie exploration o- the menu an the useronsider trade'o--s bet2een attributes.

    +enu seletion an be a ery e--etie method o- deision ma0ing 2hen the order o- theaspets re-lets their importane and there are no important interations bet2eenattributes !the presene or absene o- one attribute negates the alue o- another*.=n-ortunately, menu aided deision ma0ing an be misleading !a* i- the order o- aspetsis ill'hosen , !b* i- a 2ea0 pre-erene eliminates alternaties that hae desirable aspetson other attributes, or !* i- strong interations exist bet2een the alues o- attributes.hen sales and promotion tehni?ues are embedded in menus !e.g., menu seletion -oronline atalog shopping series*, the order o- aspets may not be in the deisionma0erGs best interest. 4or example, the shopper may 2ish to spend no more than U6& onrunning shoes. o2eer, the menu may ommit the deision ma0er to ertain brandnames or -eatures -irst and present prie only near the end o- hoie proess. At that

     point, the deision ma0er may be lured to spend muh more than U6&.

    #eeral algebrai rules hae been proposed to optimie the outome o- the deision proess. /hese rules ma0e -ull use o- the utility alues and are used in omputer aideddeision systems.

    +eighted Utilit! Rule. #elet that alternatie that has the maximum sum o- 2eightedutilities. +ultiattribute utility theory !Keeny I :ai-a, 19B

  • 8/15/2019 computer/human interface design

    66/295

    Cognitive lgebra Rules. #elet that alternatie that has the maximum alue resulting-rom an algebrai rule that simulates the deision ma0erGs Fudgment rule. "n-ormationintegration theory has demonstrated that -or di--erent types o- Fudgments people applydi--erent ognitie rules in order to ombine attribute alues to ma0e an oerallassessment !Anderson, 19&8 Norman, 19B9*.

    =i D - ! ui1, ui

  • 8/15/2019 computer/human interface design

    67/295

    di--erent ases. =sers may be searhing -or an un0no2n goal obFet that satis-ies some preset onditions !e.g., a boo0 on 2hale song*. 5n the other hand, they may be loatinga 0no2n goal obFet in order to aess in-ormation about it !e.g., the all number o-hale #ong by /ony Hohnston*. As users beome more -amiliar 2ith the system, theylearn top'leel menus and a?uire 0no2ledge about the inherent struture o- the data

     base. $ut unli0e ommand menus, users 2ill rarely, i- eer, beome -amiliar 2ith theextensie lo2er'leel menu in-ormation exept in 2ell searhed loal areas.

    "n general, menu -amiliarity is a -untion o- the -re?ueny o- aess and the userGs prior0no2ledge about obFets in the data base. "n turn, -re?ueny o- aess is a -untion o-user experiene and leel o- menu. 4igure 3.9 sho2s this relationship in terms o--amiliarity pro-iles -or inexperiened and experiened users and -or 0no2n andun0no2n target obFets as a -untion o- menu in-ormation leel. /hese pro-iles help to

     predit the amount o- time, di--iulty, and e--ort re?uired on the part o- users -oraessing in-ormation. ;esigners should realie that the greatest di--iulty 2ill be at theintermediate leels o- menu searh -or all but highly experiened users o- relatiely

    small data bases. Conse?uently, great are needs to be ta0en at this point in proidinghelp to users.

  • 8/15/2019 computer/human interface design

    68/295

    "n many ases users may hae less than a goal direted approah to data base retrieal."ndeed, they may be bro2sing the data base to gain an idea o- 2hatGs there. +enus2hih list the sets o- ategories andor -aets ?ui0ly reeal the sope o- the data base.5n the other hand, users may be pursuing a serendipitous 2al0 through the path2ays o-a data base. +enus allo2 -or a relatiely e--ortless seletion o- paths through thesystem.

    =nli0e ommand menus or deision menus, users o- in-ormation menus are drien bythe in-ormation displayed rather than ommands re?uired or alternaties desired. /his issay that users are onentrating on the output -untion o- menu seletion rather than onthe pointing, exeution, or input -untions. /o be sure many in-ormation retrieal

  • 8/15/2019 computer/human interface design

    69/295

  • 8/15/2019 computer/human interface design

    70/295

  • 8/15/2019 computer/human interface design

    71/295

  • 8/15/2019 computer/human interface design

    72/295

    0no2ledge about the -rame. /ypially the user starts 2ith either an expliitly 0no2ntarget or a partially spei-ied target. "- the target is expliitly 0no2n !4igure >.1*, theuser engages in a isual mathing proess. 4or eah alternatie sanned, the proessdetets either a math or a mismath. #ine errors an our, the lassi t2o'2ay tableo- possibilities -rom signal detetion theory !@reen I #2ets, 1966* obtains as sho2n in

    the bottom panel o- 4igure >.1. "t is generally the ase that the proessing time is -aster-or a math than -or a mismath !e.g.*. #eond, to the extent that any trans-ormation onthe stimulus is re?uired to proess a omparison, response time 2ill be inreased !e.g.*./hird, to the extent that alternaties are similar and on-usable, there 2ill be an inreasein the number o- errors !e.g., Kinney, +arsetta, I #ho2man, 1966*. +enus 2hih useisually and semantially distint alternaties 2ill result in -aster response times and-e2er error. "n pratie, ho2eer, labels are not al2ays distint and may lead toinreased proessing time and seletion errors.

  • 8/15/2019 computer/human interface design

    73/295

    "- the target is partially spei-ied, the user engages an enoding and ealuation proessas sho2n in 4igure >.

  • 8/15/2019 computer/human interface design

    74/295

    "n the ase o- partially spei-ied goals, users may either ealuate all o- the alternatiesand selet the alternatie haing the highest ealuation !le-t panel o- 4igure >..

  • 8/15/2019 computer/human interface design

    75/295

    #earh strategies are also harateried by their stopping rule. "n a sel-'terminatingsearh, the user stops 2hen the desired item is enountered. An exhaustie searhre?uires the user to inspet all o- the items prior to ma0ing a hoie. 4inally, in aredundant searh a-ter all the items hae been inspeted, the user still annot ma0e ahoie and must re'inspet some items. +enus and tas0s that promote sel-'terminating

    searh are expeted to be -aster than 2hen users must examine all items exhaustielyand redundantly. /ypially, sel-'terminating searh ours 2hen the user has anexpliitly 0no2n target in mind and need only reognie a math bet2een the target andan item. #el-'terminating searh may also our i- the subFet uses the strategy o-satis-iing. "- none o- the alternaties meet the riteria be-ore the list is exhausted, nodeision has been ahieed, and the user must adopt a di--erent strategy. "- the user 0epttra0 o- an ealuation o- eah alternatie, he or she may pi0 the alternatie haing thehighest sore. $ut more li0ely than not, the user may hae to go ba0 and re'ealuatealternaties in order to 2eigh the pros and ons assoiated 2ith items still in therunning.

    en a-ter assessing all o- the alternaties, it is possible that none o- them proessatis-atory. /he user has exhausted the list o- options and not -ound any that meet his

  • 8/15/2019 computer/human interface design

    76/295

  • 8/15/2019 computer/human interface design

    77/295

    Assuming that eah saade ta0es about the same amount o- time t , the aerage time todetet a target 2ill be # D t  p.

    "- there is one orret alternatie in a list o- n, the probability o- -inding the target on a partiular saade 2ill be p D 1n and # D nt .

    Conse?uently, searh time is again a linear -untion o- the number o- items. And thegeometri model predits the same aerage time as the Lee I +a@regor model -or anexhaustie searh. /he maFor di--erene is that the predited ariability 2ill be muhgreater in the Card model than in the Lee I +a@regor model. Lee I +a@regoremphasie reading time beause they are primarily addressing ideotext systems. Cardemphasies saade and isual searh time -or ommand menus. =n-ortunately, bothmodels ignore the deision proess and assume that hoie time !distint -rom readingor isual sanning time* does not ary 2ith the number o- alternaties. A later setion2ill address this issue.

    "t is o-ten the ase that users hae more than one possible target -or 2hih they aresearhing. #eeral di--erent items may satis-y the re?uirements o- the searh. 4orexample, the user may be loo0ing -or either JstopJ or J?uit.J An extensie series o-studies on isual and memory sanning !Neisser, 19638 #hneider I #hri--in, 19BB8#hi--rin I #hneider 19BB* sho2 the relationship bet2een the number o- possibletargets and the total response time. /he experimental tas0 is analogous to menuseletion. A subFet is as0ed to searh -or a target in a display o- haraters. 4orexample, one might be as0ed to searh the array sho2n in the upper panel o- 4igure >.>and report 2hen the target has been -ound. /he target may be simply de-ined as Jtheletter L,J or the Jletters L, + or 7.J "n general, the greater the number o- possibletargets, the longer it ta0es to detet the one that is atually there. /he upper panel o-4igure >.( sho2s the idealied results o- suh experiments using -rom one to six

     possible targets. /he results indiate that there is a linear inrease in searh time as thenumber o- possible targets inreases. Presumably subFets sanned eah item and thenompared eah o- the possible targets in the target set 2ith the item. ah omparisonadded a onstant amount o- time.

  • 8/15/2019 computer/human interface design

    78/295

    /he intriguing result o- these studies ours 2hen subFets pratie the same set o-

    targets oer an extended period o- pratie 2ee0 a-ter 2ee0 and month a-ter month. /heresults o- these studies indiate that di--erenes due to both the number o- targets andthe number items sanned derease greatly. /he lo2er panel o- 4igure >.( sho