Professional Documents
Culture Documents
MakinginOrganizations
UNIT1.2
DECISIONSINVOLVING
MULTIPLEOBJECTIVES
Wherearewenow?
Decision
Analysis
Problem
definition
Heuristics
andBiases
Choiceand
Preference
Applications
Twosystemsof
thinking
Prospecttheory
Performance
appraisal
Escalation of
commitment
Motivating
employeesby
money?Or?
Timeand
preferences
Whypeople
becomebad
apples?
Ismorechoice
alwaysbetter?
Goinggreen
Availability
heuristic
Decisionsinvolving
multipleobjectives
Decisionmaking
underuncertainty
Representativeness
heuristic
Anchoringand
adjustment
Embodied
cognition
Collectingmore
information
Bounded
awareness
UNIT1.2
DECISIONSINVOLVINGMULTIPLEOBJECTIVES
Howpeopleusuallydealwithdecisionsthatinvolvemultiple
objectives?
Lexicographicchoice
Semilexicographicchoice
Eliminationbyaspects
Sequentialdecisionmaking:Satisficing
Usingdecisionanalysistodealwithmultipleobjectivesand
tradeoff:
SimpleMultiAttributeRatingTechnique(SMART)
Goodwin,P.andWright,G.(2009).DecisionAnalysisforManagementJudgment,4th Edition.UK:JohnWiley&Sons.
Chapters23.
Toevaluatealternatives,westartby
CONSTRUCTINGTHECONSEQUENCESTABLE
Alternatives
Objectives
Weneedtospecifyhowwelleachalternativemeetseachofourobjectives.
Someobjectivesinvolvenatural measurement
Maximizeprofit:$$$
MaximizeexposureofaTVad:Numberofpeoplesurveyedwhorecall
seeingthead
Someobjectivesinvolvesubjective measurement
Jobcandidate'sabilitytofacilitatecontributionsofteammembers:
Excellent,Good,Average,orPoor
(seenextslideforanexampleofmeasurementtool)
4
Forexample:TheAssociationofAmericanCollegesandUniversities
usesthefollowingsubjectivecriteriatomeasurestudentsabilityto
facilitatecontributionofteammembers:
Excellent:Engagesteammembersinwaysthatfacilitatetheircontributions
tomeetingsbybothconstructivelybuildinguponorsynthesizingthe
contributionsofothersaswellasnoticingwhensomeoneisnot
participatingandinvitingthemtoengage.
Good:Engagesteammembersinwaysthatfacilitatetheircontributionsto
meetingsbyconstructivelybuildinguponorsynthesizingthecontributions
ofothers.
Average:Engagesteammembersinwaysthatfacilitatetheircontributions
tomeetingsbyrestatingtheviewsofotherteammembersand/orasking
questionsforclarification.
Poor:Engagesteammembersbytakingturnsandlisteningtoothers
withoutinterrupting.
Themeaningofeachcategory(e.g.excellent,poor)shouldbeclearly
described.
5
Wecollectinformationontheconsequences(howwelleach
alternativemeetseachobjective),andputthisinformationintoa
moreusefulformat:Theconsequencestable.
Example:Choosingamongjoboffers
Salary $/monthly
Flexibilityofworkschedule Low,Moderate,High
Vacationtime #ofdays/year
Enjoyment Boring,Average,Great
JobA
JobB
JobC
JobD
JobE
Monthlysalary($)
16,000
19,200
15,200
11,400
11,400
Flexibilityofwork
schedule
Moderate
Low
Moderate
High
Low
14
12
15
10
10
Great
Average
Boring
Average
Boring
Vacation(days)
Enjoyment
DOMINANCEANDTRADEOFFS
Dominatedalternatives the
alternativesperformworsethanall
otheralternativesonallattributes(or
worseonsomeattributesandequally
wellontheremainingattributes);
eliminatethem!
(Better)
Alternative1
AttributeY
Agoodstartingplacewhenyoufinishyour
consequencestableistolookfor
Dominantalternative thealternative
thatperformsbetterthanallother
alternativesonallattributes (or
performsbetterthanotheroptionson
someattributesandperformsjustas
wellontheremainingattributes);thisis
yourbestchoice!
Alternative2
(Better)
AttributeX
(Better)
Whennoalternativedominates
theothers.
Youneedtodealwithtradeoffs.
AttributeY
Alternative1
Alternative2
(Better)
AttributeX
Anydominantalternative?Anydominatedalternative?
JobA
JobB
JobC
JobD
JobE
Monthlysalary($)
16,000
19,200
15,200
11,400
11,400
Flexibilityofwork
schedule
Moderate
Low
Moderate
High
Low
14
12
15
10
10
Great
Average
Boring
Average
Boring
Vacation(days)
Enjoyment
Nowwehaveeliminatedthedominatedalternative.
Whatsnext?
WITHOUTEXPLICITLYHANDLINGTRADEOFFS,
howpeopleusuallydealwithdecisionsthatinvolvemultipleobjectives?
Decisionmakersoftenusethefollowingmethodstomakedecisions
involvingmultipleobjectives:
Lexicographicchoice
Semilexicographicchoice
Eliminationbyaspects
Sequentialdecisionmaking:Satisficing
10
LexicographicChoice
Steps:
1. Pickthemostimportantattribute.
2. Choosethealternativethathasthebestperformanceonthat
attribute.
3. Ifthereisatie,moveontothenextmostimportantattribute.
11
Supposeyourmostimportantattributeis#ofvacationdays,
2nd enjoyment,3rd flexibilityofworkschedule,and4th salary:
Usinglexicographicchoice,yourchoicewillbe:______________
JobA
JobB
JobC
JobD
Monthlysalary($)
16,000
19,200
15,200
11,400
Flexibilityofwork
schedule
Moderate
Low
Moderate
High
14
12
15
10
Great
Average
Boring
Average
Vacation(days)
Enjoyment
12
Islexicographicchoiceagoodwaytomakedecisions?
Itsquickif therearefewties.
Itseasy,itonlyrequiresyoutorankyourattributes(e.g.Issalary
moreimportantthanvacationdays?).
BUTlexicographicchoiceisnoncompensatory:
Analternativespoorperformanceononeattributecannotbe
compensatedforbygoodperformanceonotherattributes.
E.g.Ifthemostimportantattributeissalaryandvacationdayscomes
next,then$10ksalaryand1vacationday isstillpreferred to$9,999
salaryand100vacationdays.
JobX
JobY
MonthlySalary
$10,000
$9,999
Vacation
1dayperyear
100daysperyear
13
SemilexicographicChoice
Thisdiffersslightlyfromlexicographicchoiceinthat,iftheperformance
ofalternativesonanattributeissimilar,thedecisionmakerconsiders
themtobetiedandmovesontothenextattribute.
Supposeyourmostimportantattributeis#ofvacationdays,
2nd enjoyment,3rd flexibilityofworkschedule,and4th salary.ANDyour
ruleis:ifdifferencein#ofvacationdaysislessthan2,choosethejob
withgreaterenjoyment.
Yourchoice:_______
JobA
JobB
JobC
JobD
Monthlysalary($)
16,000
19,200
15,200
11,400
Flexibilityofwork
schedule
Moderate
Low
Moderate
High
14
12
15
10
Great
Average
Boring
Average
Vacation(days)
Enjoyment
14
EliminationbyAspects
Steps:
1. Pickthemostimportantattribute.
2. Decideonacutoffpointforthatattribute.
3. Eliminateallalternativesthatdonotmeetthatcutoff.
4. Moveontothenextmostimportantattribute.
15
Supposeyourmostimportantattributeissalary(cutoff:$15,000)
2nd :vacation(cutoff:12days)
3rd :flexibilityofworkschedule(cutoff:moderate)
4th :enjoyment(cutoff:average)
Usingeliminationbyaspects,yourchoicewillbe:______________
JobA
JobB
JobC
JobD
Monthlysalary($)
16,000
19,200
15,200
11,400
Flexibilityofwork
schedule
Moderate
Low
Moderate
High
14
12
15
10
Great
Average
Boring
Average
Vacation(days)
Enjoyment
16
Iseliminationbyaspectsagoodwaytomakedecisions?
Itsquickandeasy.
Itprotectsyoufrommistakenlyselectinginferioralternatives
becausealternativesthatdonotmeetthecutoffareremoved.
Itisalsoagoodstrategytonarrowdownahugenumberof
alternativestoafewfinaliststhatyoucanevaluateinmoredetail.
But,itsalsononcompensatory:
Onlyoneattributeisconsideredatatime,notexplicitlyhandling
tradeoffsamongattributes.
17
SequentialDecisionMaking:Satisficing
Thepreviousmethodsareintendedtodescribehowpeoplemakea
decisionwhentheyevaluateallalternativessimultaneously.
Insomesituations,however,decisionmakersevaluatealternatives
sequentially.
Peoplesatisfice,i.e.searchfornewalternativesuntiltheyfinda
satisfactoryalternativethatwillsuffice.
Thusthechosenoptionisjustgoodenough,maynotbethebest
availableoption.Thechoicealsodependsontheorderinwhichthe
alternativespresentthemselves.
18
HANDLINGTRADEOFFS:
SIMPLEMULTIATTRIBUTERATINGTECHNIQUE(SMART)
Steps:
1. Foreachattribute,assignsubjectivevalues(utilities)
toreflecttheperformanceofthealternativesonthatattribute.
2. Determineaweightforeachattribute.
3. Foreachalternative,computeaweightedutility.
Then,makeaprovisionaldecision.
4. Performsensitivityanalysis.
Letsillustratewithanexample.
19
Step1:Foreachattribute,assignsubjectivevalues(utilities)toreflect
thealternativesperformanceonthatattribute.
Foreachattribute,identifythemostpreferredlevelandassignita
utilityof 1,thenidentifytheleastpreferred level andassignita
utilityof0.
Examineeachintermediate levelandassignitautilityinbetween0
and1 sothatthespacebetweenutilitiesrepresentsthestrengthof
preferenceforoneattributeleveloveranother.
Mostpreferred
Leastpreferred
Flexibilityofwork
schedule
JobA
JobB
JobC
JobD
Moderate
(?)
Low
(0.0)
Moderate
(?)
High
(1.0)
Intermediatelevel
20
E.g.Iftheimprovementfrom
0.75
Utility
lowtomoderateisseenby
youtobe50%aspreferableas
theimprovementfromlowto
high,assignmoderateautility
of0.5.
0.5
0.25
0
Low
Moderate
High
E.g.Iftheimprovementfrom
0.75
Utility
lowtomoderateisseenby
youtobe75%aspreferableas
theimprovementfromlowto
high,assignmoderateautility
of0.75.
0.5
0.25
0
Low
Rememberthatitistheinterval (or
Moderate
High
improvement)betweentheattribute
levelsthatwecompare.
21
JobA
JobB
JobC
JobD
Monthlysalary($)
16,000
(0.7)
19,200
(1.0)
15,200
(0.6)
11,400
(0.0)
Flexibilityofwork
schedule
Moderate
(0.5)
Low
(0.0)
Moderate
(0.5)
High
(1.0)
Vacation(days)
14
(0.6)
12
(0.4)
15
(1.0)
10
(0.0)
Enjoyment
Great
(1.0)
Average
(0.7)
Boring
(0.0)
Average
(0.7)
22
Step2:Determineaweightforeachattribute.
Imaginetwohypotheticalalternatives:onewiththeworst
performanceonallattributes,anotherwiththebest.
Foreachattribute,considertheimportanceofthechangefrom
theworstleveltothebestlevel.
Fortheattributethatyouwouldmostliketomovefrom
theworstleveltothebestlevel,assign aweightof100.
Assignweightstotherestofattributesbasedontheirimportance
relativetothemostimportantchange.
Normalizetheweights.
The weights reflect the importance of
changes (or swings) not the
importance of attributes per se
23
JobA
JobB
JobC
JobD
Monthlysalary($)
16,000
(0.7)
19,200
(1.0)
15,200
(0.6)
11,400
(0.0)
Flexibilityofwork
schedule
Moderate
(0.5)
Low
(0.0)
Moderate
(0.5)
High
(1.0)
Vacation(days)
14
(0.6)
12
(0.4)
15
(1.0)
10
(0.0)
Enjoyment
Great
(1.0)
Average
(0.7)
Boring
(0.0)
Average
(0.7)
Hypothetical
worst
alternative
Hypothetical
best
alternative
Monthlysalary($)
11,400
19,200
Flexibilityofwork
schedule
Low
High
10
15
Boring
Great
Vacation(days)
Enjoyment
Swing
Weight
24
Inthiscase,theswingfromBoringtoGreatenjoymentisconsidered
tobe50%asimportantastheswingfrom$11,400to$19,200monthly
salary.
Monthlysalary($)
Flexibilityofwork
schedule
Vacation(days)
Enjoyment
Hypothetical
worst
alternative
Hypothetical
best
alternative
Swing
Weight
Normalized
Weight
11,400
19,200
100
100/180=
0.556
Low
High
20
20/180=
0.111
10
15
10
10/180=
0.056
Boring
Great
50
50/180=
0.278
Total
180
25
Whyswingweights?
Supposethatthechoiceisbetweenjusttwojobsandthatweare
evaluatingtheseononlytwoattributes.
Ifweconsiderrelativeimportanceofattributes,say,youconsider
salarytobefivetimesmoreimportantthanvacationdays
JobX
JobY
MonthlySalary
(5)
$10,002
(1.0)
$10,000
(0.0)
Vacation
(1)
10daysperyear
(0.0)
20daysperyear
(1.0)
AggregatedutilityofJobX=5,thatofJobY=1;butsacrificing10days
ofvacationtogain$2soundsstrange.
Ifthealternativesperformverysimilarlyonaparticularattribute (i.e.
therangebetweenworstandbestissmall),thenthisattributeis
unlikelytobeimportantinthedecision,eventhoughthedecisionmaker
mayconsiderittobeanimportantattributeingeneral.
26
Step3:Foreachalternative,computeaweightedutility.Then,
makeaprovisionaldecision.
JobA
JobB
JobC
JobD
Monthlysalary($)
[0.556]
16,000
(0.7)
19,200
(1.0)
15,200
(0.6)
11,400
(0.0)
Flexibilityofwork
schedule[0.111]
Moderate
(0.5)
Low
(0.0)
Moderate
(0.5)
High
(1.0)
Vacation(days)
[0.056]
14
(0.6)
12
(0.4)
15
(1.0)
10
(0.0)
Enjoyment
[0.278]
Great
(1.0)
Average
(0.7)
Boring
(0.0)
Average
(0.7)
Weighted utilityof
eachalternative
0.756
0.772
0.444
0.306
27
Step4:Performsensitivityanalysis.
EventhoughourSMARTapproachisverysystematic,itstilldepends
uponsubjectivejudgments.
Howgoodaretheintermediatelevelsofattributeperformance
(0 utility 1)
Howimportantaretheattributesinthedecision
(0 weight 100,beforenormalizing)
Thepurposeofsensitivityanalysisistodeterminehowsensitiveyour
choicesaretothejudgmentsthatyouvemade.Inotherwords,if
yourjudgmentsweresomewhatdifferent,wouldyouchoiceremain
thesame?
Onjudgmentsthatyouarenotveryconfidentabout,pleaseconduct
asensitivityanalysis.
28
Example:
Previously,weassignedtheenjoymentattributeaweightof50.
Willourchoiceremainthesameifthisattributeisweighteddifferently?
Whichalternativeisthebestwhenenjoyment hasaweightof0?
Whichalternativeisthebestwhenenjoyment hasaweightof100?
Hypothetical
worst
alternative
Hypothetical
best
alternative
Swing
Weight
Monthlysalary($)
11,400
19,200
100
Flexibilityofwork
schedule
Low
High
20
10
15
10
Boring
Great
50
Vacation(days)
Enjoyment
29
Whenenjoyment hasaweightof0:
JobA
JobB
JobC
JobD
Monthlysalary($)
[100/130 =0.77]
16,000
(0.7)
19,200
(1.0)
15,200
(0.6)
11,400
(0.0)
Flexibilityofwork
schedule
[20/130 =0.15]
Moderate
(0.5)
Low
(0.0)
Moderate
(0.5)
High
(1.0)
Vacation(days)
[10/130 =0.08]
14
(0.6)
12
(0.4)
15
(1.0)
10
(0.0)
Enjoyment
[0/130 =0.00]
Great
(1.0)
Average
(0.7)
Boring
(0.0)
Average
(0.7)
Weighted utilityof
eachalternative
0.662
0.800
0.615
0.154
30
Whenenjoyment hasaweightof100:
JobA
JobB
JobC
JobD
Monthlysalary($)
[100/230 =0.43]
16,000
(0.7)
19,200
(1.0)
15,200
(0.6)
11,400
(0.0)
Flexibilityofwork
schedule
[20/230 =0.09]
Moderate
(0.5)
Low
(0.0)
Moderate
(0.5)
High
(1.0)
Vacation(days)
[10/230 =0.04]
14
(0.6)
12
(0.4)
15
(1.0)
10
(0.0)
Enjoyment
[100/230 =0.43]
Great
(1.0)
Average
(0.7)
Boring
(0.0)
Average
(0.7)
Weighted utilityof
eachalternative
0.809
0.757
0.348
0.391
31
Ourchoiceissensitivetotheweightplacedonenjoyment.
ItcanbeseenthatJobBisthebestalternativeaslongastheweight
placedonenjoymentislessthan60.
WeightedutilityofAlternatives
0.9
0.8
JobB
0.8
0.7
JobA
0.662
0.615
0.6
0.809
0.757
0.5
JobC
0.4
0.391
0.348
0.3
JobD
0.2
0.154
0.1
0
0
10
20
30
40
50
60
WeightonEnjoyment
70
80
90
*YoumayrefertotheExcelfileonCanvasformoreinformationonhowtoconstructthisgraph.
100
32
Carryingoutsensitivityanalysismightleaddecisionmakersto
reconsidersomeofthefigurestheysupplied.Butinmanycases,
sensitiveanalysisshowsthatourchoicesarestableoverarangeof
thosefigures.
Notethatsensitivityanalysisonlyallowsdecisionmakersto
investigatetheeffectofchangingone variable(valueorweight)at
atime.
Mustajoki etal.(2006)havedevelopedamodelingmethodto
simultaneouslyvaryweightsandvalues.Takealookonlyifyouare
interestedandgoodatmathematics.
Mustajoki,J.,Hamalainen,R.P.,andLindstedt,M.R.K.(2006).Usingintervalsforglobal
sensitivityandworstcaseanalysesinmultiattribute valuetrees.EuropeanJournalof
OperationalResearch,174,278292.
33
Inclassexercise:
Previously,weassignedmoderatelevelofflexibilityautilityof
0.5.
Isourchoicesensitivetotheutilityassignedtomoderate
flexibility?
Hints:Calculatetheweightedutilitiesofthealternativeswhen
moderateflexibilityisassigned0and1.
JobA
JobB
JobC
JobD
Monthlysalary($)
[0.556]
16,000
(0.7)
19,200
(1.0)
15,200
(0.6)
11,400
(0.0)
Flexibilityofwork
schedule[0.111]
Moderate
(0.5)
Low
(0.0)
Moderate
(0.5)
High
(1.0)
Vacation(days)
[0.056]
14
(0.6)
12
(0.4)
15
(1.0)
10
(0.0)
Enjoyment
[0.278]
Great
(1.0)
Average
(0.7)
Boring
(0.0)
Average
(0.7)
34
Youranswerhere:
Whenmoderateflexibilityhasautilityof0,theweightedutilitiesare:
JobA:______ JobB:______ JobC:______ JobD:______
Thusthechoiceshouldbe________.
Whenmoderateflexibilityhasautilityof1,theweightedutilitiesare:
JobA:______ JobB:______ JobC:______ JobD:______
Thusthechoiceshouldbe________.
Ourchoiceis(sensitive/insensitive)toutilityassignedtomoderate
flexibility.
35
Conflictsbetweenintuitiveandanalyticresults
Yourintuitivechoicemaybedifferentfromthatobtainedthrough
SMART.Thisconflictshouldbeexamined.
Thisconflictcouldbebecausetheproblemwastoolargeand
complexforyoutohandlewithoutusingDA,sothatyourtrue
preferenceswerenotreflectedinyourintuitivejudgment.
Researchhassuggestedthatthecorrelationbetweenpreference
derivedfromintuitivejudgmentandpreferencederivedfrom
SMARTdecreasesasthenumberofattributesintheproblemgets
larger.
Theconflictmayalsosuggestthatanimportantelementofthe
problemhasnotbeencapturedbytheanalysis(e.g.missingor
overlappingobjectives).
36
AnExampleofDecisionMakingUsingSMART
Trainingonresearch
(Proxy:Research ranking)
Trainingonteaching
Commutetimeandmeans
Doing a PhD in
Management
HKUST
HKU
HKBU
Ranked6
Ranked77
(1.0)
(0.3)
Notin thetop
100
(0.0)
Poor
Poor
Good
(0.0)
(0.0)
(1.0)
35 minutesby
taxi
(0.0)
5 minutesby
walk
(1.0)
20minutesby
taxi
(0.2)
Foreachattribute,
Themostpreferredlevelhasautilityof1
Theleastpreferredlevelhasautilityof0
Intermediateoneshaveautilityinbetween0and1
(i.e.0 utility 1)
37
Hypotheticalworst Hypotheticalbest
alternative
alternative
Trainingonresearch
(Proxy:Research ranking)
Trainingonteaching
Commutetimeandmeans
Swing Weight
Notin thetop100
Ranked6
100
Poor
Good
80
35minutesbytaxi
5minutesbywalk
20
Foreachattribute,decidehow importantitisforyouto
movefromtheworstoutcometothebestoutcome
Giveaweightof100tothemostimportantmove
Othermoves,giveaweightinbetween0and100
(i.e.0 weight 100)
38
Normalizetheweights
Trainingonresearch
(Proxy:Research ranking)
(0.5)
Trainingonteaching
(0.4)
Commutetimeandmeans
(0.1)
Weighted utility
HKUST
HKU
HKBU
Ranked6
Ranked77
(1.0)
(0.3)
Notin thetop
100
(0.0)
Poor
Poor
Good
(0.0)
(0.0)
(1.0)
35 minutesby
taxi
(0.0)
0.50
5 minutesby
walk
(1.0)
20minutesby
taxi
(0.2)
0.25
0.42
PickthealternativewiththeHIGHESTweighted
utilityastheprovisionalchoice
39
Sheisnotsosureabouttheweightshegavetoteachingrelatedtraining,
sosheconductsasensitivityanalysis.
SensitivityanalysisshowsthatHKUSTisthebestalternativeaslongasthe
weightplacedonTeachingrelatedtrainingislessthan96.
Weightedutilityofalternatives
0.90
0.80
0.70
0.60
0.50
HKUST
0.40
HKU
0.30
HKBU
0.20
0.10
0.00
0
10
20
30
40
50
60
70
80
Weightonteachingrelatedtraining
90
100
40
TimeforREFLECTION
Describeadecisionproblemthatyouretryingtosolve.
Howwouldyoucollectinformationabouttheconsequences?
Whatdecisionstrategywouldyouuse(lexicographicchoice,elimination
byaspects,SMART,etc.)?Why?
41
INTENDEDLEARNINGOUTCOMESFORUNIT1.2
Bynow,youshouldbeableto:
Applythefollowingdecisionstrategiesanddescribetheir
prosandcons
Lexicographicchoice
Semilexicographicchoice
Eliminationbyaspects
Sequentialdecisionmaking:Satisficing
PerformaSMARTdecisionanalysis(besurethatyou
understandeverystep)
42