testing boundaries: a theory of adaptation and framing ......university of bath phd testing...
TRANSCRIPT
University of Bath
PHD
Testing Boundaries: A Theory of Adaption and Framing Effects in Ongoing Tasks
Harrison, Timothy
Award date:2012
Awarding institution:University of Bath
Link to publication
Alternative formatsIf you require this document in an alternative format, please contact:[email protected]
General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal ?
Take down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.
Download date: 07. Nov. 2020
1
TestingBoundaries:ATheoryofAdaptionandFramingEffectsinOngoing
Tasks
TimothySamuelHarrison
AthesissubmittedforthedegreeofDoctorofPhilosophy
UniversityofBath
DepartmentofComputerScience
May2012
COPYRIGHTAttention isdrawnto the fact thatcopyrightof this thesis restswiththeauthor.Acopyof this thesishasbeensuppliedoncondition thatanyonewho consults it is understood to recognise that its copyrightrestswith theauthorand that theymustnot copy itorusematerialfromitexceptaspermittedbylaworwiththeconsentoftheauthor.
This thesis may be made available for consultation within theUniversityLibraryandmaybephotocopiedor lent toother librariesforthepurposesofconsultation.
2
3
Abstract
Thisthesis investigateshowinformationpresentationaffectsdecisions in ongoing task scenarios. For this purpose it re‐applies theprinciplesofbounded rationality and specificallyframing effects into this domain. Over a number of studies,unique properties concerning both frame effectiveness andadditionalmeasuressuchasconfidenceareobservedtooccur.Atheoryofcognitiveadaptationtonovelscenarios,andare‐defining of the concept of framing effects are proposed as aresult.
4
5
Acknowledgements
It isaclichéthataPhDthesisowes itsexistencetomorethan justoneperson.Butitisaclichébecauseitistrue,andthisthesisisnoexceptiontothatrule.
Iwish to thankmyparents, for being supportivewithout question throughoutthis process with its many ups and downs. And more than that, for theunconditional love,supportandcommitmentthatstartedwhenIwasbornandhasn’tfalteredsince.
IwouldalsoliketothankthemanyfineteachersIhavebeenprivilegedtolearnfrom inmytime,both inareas familiar to this thesisandmanyothersbesides.MrsShepard,PeterChapman,MrEnnis,MrHolleleyandcountlessothers,yourlessonsinspired,andsomehowtheyledmehere.
Finally I would like to thank my supervisor Peter Johnson. For guiding theprocesswhereitwasnecessary,butalsoforsteppingbackandforcingmetogofindmyownanswers–andquestions–whenappropriate.Theworkpresentedin thisdocument isonesideofmyPhDexperience,but theprocessbywhich Iarrivedatitisequallysignificanttome.Peterguidedboth,andhasmydeepandsincerethanksasaresult.
6
7
TableofContents
CHAPTERONEDecisions,Decisions,Decisions ................................................. 10WhyDecisions? .........................................................................................................................10TheScienceofDecisionMaking ..........................................................................................14Structure .....................................................................................................................................15
CHAPTERTWOLiteraturereview .......................................................................... 17Categorization ...........................................................................................................................17AssumptionsandBiasing ......................................................................................................18StrategyandPlanning ............................................................................................................20SituatedCognition&ProblemIsomorphs .......................................................................22BoundedRationality ...............................................................................................................24SystemsOneandTwo .........................................................................................................................25ProspectandUtilitytheory...............................................................................................................26FramingEffects ......................................................................................................................................29LimitationsofFraming .......................................................................................................................33Re‐definingFraming ............................................................................................................................34
CHAPTERTHREE–EquivalentRuleChanges:TheSolitaireStudy................ 37Introduction...............................................................................................................................37Design ..........................................................................................................................................37Method.......................................................................................................................................................38Solitaire .....................................................................................................................................................39SortingandCoding ...............................................................................................................................42Participants..............................................................................................................................................43
ResultsandAnalysis ...............................................................................................................43OverallData .............................................................................................................................................43ConfidenceandPerformance...........................................................................................................44LearningandProgression .................................................................................................................45PatternAnalysis.....................................................................................................................................48
AModelofNovelAdaptation................................................................................................52Explanations&Predictions...............................................................................................................54
Discussion...................................................................................................................................55
CHAPTERFOUR–FramingTaskInstructions:TheCardGameStudy .......... 56Introduction...............................................................................................................................56Design ..........................................................................................................................................56GameRules...............................................................................................................................................57FrameConstruction .............................................................................................................................58Hypotheses...............................................................................................................................................63ExperimentalProcedure ....................................................................................................................63Caveats.......................................................................................................................................................65CollectionandAnalysis.......................................................................................................................66QuestionnaireScales............................................................................................................................67Participants..............................................................................................................................................67
Results .........................................................................................................................................67OverallScore ...........................................................................................................................................67PlayingTimes..........................................................................................................................................68ContextualData......................................................................................................................................69AceData.....................................................................................................................................................71Discarding ................................................................................................................................................75
8
Discussion...................................................................................................................................76HypothesisTesting ...............................................................................................................................76AdditionalConclusions.......................................................................................................................77ModelImplications...............................................................................................................................82Significance ..............................................................................................................................................83NewHypotheses ....................................................................................................................................85
CHAPTERFIVE–FramePositioningandConfidenceinaDescriptiveTask:TheFestivalStudy .......................................................................................................... 87Introduction...............................................................................................................................87Design ..........................................................................................................................................87Overview...................................................................................................................................................87Questions ..................................................................................................................................................88FramingEffects ......................................................................................................................................89Feedback ...................................................................................................................................................89Programming ..........................................................................................................................................90Pretesting .................................................................................................................................................91Grouping&Participants.....................................................................................................................92Hypotheses...............................................................................................................................................92
Results .........................................................................................................................................92PretestingMeasures ............................................................................................................................92Decisions...................................................................................................................................................93Statisticalnote ........................................................................................................................................96ConfidenceData .....................................................................................................................................97TimingData .............................................................................................................................................99
Discussion................................................................................................................................ 104Hypotheses............................................................................................................................................ 104Complexity ............................................................................................................................................ 105FramingPositionEffects ................................................................................................................. 106Integration ............................................................................................................................................ 107Confidence............................................................................................................................................. 110ImplicationsfortheModel ............................................................................................................. 112NewQuestions..................................................................................................................................... 112
CHAPTERSIX–FramePersistenceandCyberInfluenceasaFramingMechanism:TheHerbalStudy .................................................................................114Introduction............................................................................................................................ 114Setting ..................................................................................................................................................... 114
Design ....................................................................................................................................... 115Method.................................................................................................................................................... 115Hypotheses............................................................................................................................................ 121
Results ...................................................................................................................................... 121PretestingData.................................................................................................................................... 121Choices.................................................................................................................................................... 123Confidence............................................................................................................................................. 127TimingData .......................................................................................................................................... 131RethinkData......................................................................................................................................... 140Extremists ............................................................................................................................................. 143
Discussion................................................................................................................................ 147Approach................................................................................................................................................ 147HypothesisTesting ............................................................................................................................ 147BroaderThemes ................................................................................................................................. 151
CHAPTERSEVEN–AdaptationandFramesinTasks:DiscussionAndSpeculation .....................................................................................................................162
9
Introduction............................................................................................................................ 162Approach................................................................................................................................................ 163
TheModelofAdaptation .................................................................................................... 164Integration ............................................................................................................................................ 166Formulation.......................................................................................................................................... 172
ReboundedRationality ....................................................................................................... 175RationalityinTasks........................................................................................................................... 177Applications.......................................................................................................................................... 179
FurtherWork.......................................................................................................................... 185AFinalWord ........................................................................................................................... 188
References ......................................................................................................................190
AdditionalAcknowledgements................................................................................195
Appendices……………………………………………………………………………………………197One:SolitaireStudy‐Pre‐StudyQuestionnaire.................................................................... 198Two:SolitaireStudy‐EndQuestions ........................................................................................ 199Three:CardGameStudy–Pre‐StudyQuestionnaire.......................................................... 200Four:CardGameStudy–Post‐StudyQuestionnaire .......................................................... 202Five:RiskTaking/ImpulsivenessQuestionnaire ............................................................... 203Six:F‐ScaleAuthoritarianismQuestionnaire......................................................................... 205Seven:Introversion/ExtraversionQuestionnaire ............................................................... 206Eight:FestivalStudyInstructionsForNeutralandQuestionFrameconditions.... 207Ten:FestivalStudyOver‐FrameInstructions,Internal..................................................... 209Eleven:FestivalStudySampleDecision,NeutralandOver‐FrameConditions ...... 210Twelve:FestivalStudySampleDecision,QuestionFrameInternal............................. 211Thirteen:FestivalStudySampleDecision,QuestionFrameExternal ......................... 212Fourteen:FestivalStudyQualitativeFeedbackPatternandValues ............................ 213Fifteen:FestivalStudyNumericFeedbackPatternandValues ..................................... 214Sixteen:FestivalStudyDecisionPageScreenshot............................................................... 215Seventeen:FestivalStudyFeedbackPageScreenshot....................................................... 215Eighteen:HerbalStudyAlternativeTherapiesattitudequestionnaire ...................... 216Nineteen:HerbalStudySocialMediaQuestionnaire.......................................................... 217Twenty:HerbalStudyInstructions ............................................................................................ 218Twenty‐One:HerbalStudyFeedbackPatternandValues ............................................... 220Twenty‐Two:HerbalStudysampledecision ......................................................................... 222Twenty‐Three:HerbalStudySampleTweets,NeutralDecision2 ............................... 223Twenty‐Six:HerbalStudySampleTweets,Pro‐AstorDecision2 ................................. 223Twenty‐Four:HerbalStudySampleTweets,Pro‐QuetiaDecision2 ........................... 224Twenty‐Six:HerbalStudyEventText ....................................................................................... 226
10
CHAPTERONE‐Decisions,Decisions,Decisions
Afoxandascorpionmetatthebankofariver.“Carrymeacross”,said
theScorpion.“No”saidthefox,“Becauseyouareascorpion;ifIcarry
you,youwillstingmeandIwilldrown”.“ButwhywouldIdothat?”
askedthescorpion,“Wewouldbeinthemiddleoftheriver,andI
woulddrownalso.”
Convincedbythisargumentthatthescorpionwouldnotharmhim,the
foxletthesmallercreatureonhisbackandstartedacrosstheriver.At
themidwaypointhowever,thescorpionstunghim.Aspoisonfilledhis
veinsandhestartedtodrownhemanagedtoask,“Why?Nowweare
bothdead!”
“Becauseitisinmynature,”theScorpionreplied.
WhyDecisions?
Thechoiceswemake,andthereasonswemakethemhavefascinatedpeople
sincethestartofrecordedhistory.Intheabovefable,choiceisseenas
predetermined;weactnotbecausewechooseto,butbecausewearedefinedby
whatweareandmustactaccordingly.Converselytheheroesofliterature,stage
andscreenareperpetuallystrugglingtomaketherightchoice,toactagainst
theirnatureanddefywhatseemstobepredestined–fromRomeoandJuliet
defyingtheexpectationsoftheirfamiliesandpayingthecost,throughtoHarry
PotterseekingtodenytheprophesythathewoulddieatthehandsofVoldemort.
Thestruggletodetermineourownlivesthroughactionswedecideuponisa
recurringthemeofhumanexistence.
Decisionsdefineourlives.Wherewego,whowetalkto,whatweeat,drink,read
andwatch.Evenwhenwecomplainthatourlivesaredecidedforus,thatwe
havenochoiceweareatthatmomentchoosingtovoicethatfrustration.Andwe
arechoosingthepaththatseemsimmutable,toliveuptoresponsibilitiesasa
parentortonotstartourownbusiness.Asbadanideaasthealternativemight
be,westillhavetheoptionofpursuingit.Ifwedonothingelseinaday’sspanwe
11
chooseananswertothatmostfundamentalofquestions:whetherwecontinue
onlivingordecidetoenditall.Itisacommoncomplaintthatwehavenochoice,
butthisisrarelytrueinthestrictestofsenses‐havingnogoodchoices,oronly
oneviableoptionisnotthesameashavingnochoicetomakeatall.
Andsocontinuallythroughourexistencewemakethedistinctionsand
divinationsthatdefineourpathway.Evenifwewalkthepaththatathousand
havetrodbeforeus,wefindourownroutethroughthatfamiliarlandscape,or
wechoosenottodeviateandallowpreviousexperiencetoguideusforwards.
Theactofmakingdecisionsshouldbeasfamiliartousasbreathing,andyet
despiteourlifetimeofexperiencewithconstantchoicesweremainfascinatedby
them.Wereadautobiographiesbyfamouspeopletotrytoanswerthequestion
ofhowtheygottowheretheyare,andwhatdecisionsboughtthemtothatpoint.
Wedemandinterviewswithfiguresofpublicnote,tounderstandtheir
motivations.ItwasnotenoughfortheworldtoknowthatTigerWoodshad
cheatedonhiswife;wehadtoknowwhyhewoulddoit.Endlessarticles
speculatedonthisseeminglyinexplicablechoicethathadnodirecteffectonthe
livesofmostofpeopletryingtounderstandit.Whenactsofunspeakableevilare
perpetratedandterroristskillinnocentcivilians,orgunmenattackstudentsin
theirclassrooms,werepeatedlyaskwhy,whatcouldmotivatesuchanact.Entire
worksofliteraturearecenteredarounddifficultchoices,ofmoraldilemmasand
howtheyareresolvedoraddressed.Evenwhenthedecisionsareobjectionable,
whentheoutcomeishorrendousweseektounderstandwhatcoulddrivea
persontomakethosechoices,tobringtheirlifetothepointwherethatwasa
plausiblerouteforwards.
Andthisinterestisnotsimplyinternal,ortheoretical.Peoplehavebeen
interestedintheideaofartificiallifeandthoughtthroughtheages–fromthe
golemofJewishfolkloretotherobotsofsciencefiction(Asimov,1950),the
questionofanartificialintelligencehasoftenbeenusedasameanstoholdupa
mirrortoourselves.Inresearchithasoftenbeentheotherwayaround–where
anyartificialsystemsareinevitablyjudgedagainsttheeffortlessandseemingly
simplewaynormalpeoplecanmakedecisionsandjudgments.Theseproperties
12
stillescapeourgeneralabilitytoevenexplainletalonereplicateinartificial
systems.
Ourunderstandingofourownmotivationsanddecisionmakingisoften
frustratinglypoor.Mostpeopleknowthefeelingofaskingthemselves‘whydidI
saythat?’afteranexchangehasgonebadly,andnothavinganygoodanswerfor
it.Wemakeanynumberofdecisionsinadaywithnogoodreasontojustifyit.If
apersonisaskedwhytheyhadacertainfoodforlunch,theanswerwillalmost
inevitablybesomethingalongthelinesof‘becauseIfanciedit’whichisa
tautologicalnon‐answer.PeoplemakeNewYearsresolutionsintheirmillions,
andacommonexperiencesharedbyalmostallisthatoffailingtokeepthem.We
knowthatwewanttoliveadifferentwayandwechooseonalargerscaletodo
so,butthenfailtoupholdthatdecisiononaday‐to‐daybasis.Why?
Itisacommonlyheldbeliefthatitislogic–formal,rational,anempirically
justifiableexplanationofreality‐thatdrivesourchoices.Wechoosetodo
somethingbecause,logically,itiswhatweshouldbedoing.Itisthebestwayto
achieveourgoals,themostlikelywaytoattainafavourableoutcomeandweare
driventomakerational,reasonablediscriminationsbetweentheavailable
options.Mostpeopleliketothinkofthemselvesinthiswaycertainly,andwhen
askedtodefendthechoicesmade,theywillgenerallyrespondalongthoselines–
‘itseemedliketherightthingtodo’.Butofcourse,weareanythingbutlogicalin
ourchoicesandlifedecisions.Anybodywhohaseverbeenromanticallyinvolved
inanywaycantestifytothemoney,timeandeffortthatisspentinthepursuitof
somethingthatisoftenunattainableforanynumberofreasons,orevenan
activelybadanddestructiveidea.Wefrequentlymakechoicesnotbecausethey
arelogicalbutbecausewebelievethemtobe‘right’;becausewevaluekeeping
ourword,orthesanctityofhumanlife.Weevenmakedecisionsfornological
reasonatall–a‘hunch’basedoncircumstantial,spuriousevidence.Manygreat
detectivestoriesplayheavilyonthisfamiliarsensation,withalonedetective
followingtheirgutinstinctwhenalllogicwouldinsistotherwise.Butsuch
decisionmakingishardlyrestrictedtothepulpthriller–ifyouaskmostpeople
howtheypicklotterynumberstheywillanswerwithsomevariationof
‘children’sbirthdays,significantdates,age,houseIgrewupin…’despitethefact
13
thatthismethodologyhasnogreateracuityorlikelihoodofsuccessthanrandom
selection.
Althoughlogicalthinkingisoftenseenas‘correct’andwhatweshoulddesireto
perpetuate,thereareactuallysignificantdrawbacksthatmaketheabilitytoskip
purelylogicalreasoningbeneficial.Purelogicisfrequentlyinefficientandtime
consuming,nottomentionpotentiallyhazardous.Ifapersongetssickafter
eatingaparticularfoodtheywilloftenavoidthatfoodinfuture–eventhough
logicallytheydonotknowifitwasthatparticularitemorthefoodstuffin
generalthatcausedtheillness.Butinpracticalterms,theriskofgettingsick
againisenoughofadeterrentthatthecostoftestingthehypothesislogicallyis
notworththepricepaid.Inimportantdecisions,timeisoftenatapremium,and
unknownandunreliableinformationprovesabarriertoanypurelylogical
solutions–hereguesses,heuristicsandrulesofthumbcomeintotheirown.
Commonlywefallbackoncheapanddirtymeansofgettingtoanapproximate
answerthatisgoodenoughtosolvethesituationathandandmoveon.Inagame
offootballaplayermightnotknowexactlywherehispasswillendup,orwhere
theotherplayerswillbewhenitarrivesthere,buthewillmakeitwhenhehas
spaceandtheneed,becausethealternativeisgettingtackledandlosing
possession.Itisthissortofthinking–onthefly,inthemiddleofastrategictask
whereinformationisbothplentifulandsparse,andachoiceisinformedbyboth
whathasgonebeforeandisyettocome.Howisitthatwecanpickupthebasics
ofagamesimplyfromwatchingitplayed?Orstarttoadoptthecorrectnormsin
anunfamiliarsocietyunconsciously?Doingsomethingwehaveexperienceofis
onething,buthowdoweapproachthosetasksthatwehavenopriorknowledge
ofandyetmanagetocorrectlyapplyexperiencethatissomewhatrelated?
Logicalreasoningiseasytoexplain,understandandstudy,relativelyspeaking.
Therearecertainlycomplexitiestoit(manyofthem)butthefactthatitworks
throughsolidrulesandmetricshelpstodefineitasamethodologythatcanself‐
evidentlybeexpandedonitsownterms.Guesses,hunchesandheuristicsare
harder.Thereisnotasingletermtodescribewhattheycover,andhowandwhy
theconclusionstheyofferarearrivedatarehardertoquantify.Theyvary,by
definition,byperson,experience,context,andothers.Buttheyarealsopowerful
14
andhelptodefinetheabilitiesthatmakepeoplesogoodatchoosing,adapting
andadvancinginnewandnoveltasks.Complicatedastheprocessmaybe,years
ofpreviousresearchintoitandsimilarproblemssuggestthatitcanbe
comprehended,thatelementsofthisunderstandingcanbemappedanddefined
andputtogreateruse.Inparticular,wecanstarttounderstandwhatelementsof
anewtaskwecangatherfromcontext,andhowthatcontextshapesour
approachtoit.
TheScienceofDecisionMaking
Foraconceptandatopicthatisclearlysofundamentaltoourexistenceand
lacedwithsomanyinterestingquestionsandissues,decisionmakingisatopic
thatpresentsanumberofchallengestoanyattempttostudyitscientifically.The
precedingsectionillustratedthecomplexityofthearea,anditisinthis
complexitythatthescientificproblemsemerge.
Fromatop‐downviewpoint,reductionistapproachesinevitablyrepresenta
simplificationofthatcomplexity.Tryingtonarrowdowntheelementsof
decisionmakingtotheirconstituentpartslosestheinteractivityofthosevery
partsandremovestheveryphenomenathatwereunderstudy.Initialworkin
theareaofArtificialIntelligence(AI)showedthelimitationsofthissortof
approach.Itwasassumedthatsimplyrule‐basedmodellingcouldeventually
accountforallbehaviour,butitwasneverborneout(Dreyfus,1972).Similar
problemswerefoundinpsychologywhenthelimitsofbehaviourismin
explainingcomplexbehaviourwereexposed,mostnotablyintheareaof
linguistics(Chomsky,1959).
Conversely,however,bottom‐upapproacheshaveadifferentsetofproblems.
Althoughtheypresentanobviousplacefortheindividualunitsofcognitiontobe
modeled–intheneuralnetthatmakesupthebrainandpresumablygenerates
themind–howitworksremainsfundamentallyoutofreachformodern
techniques.Thebasicinteractingparametersofthebrainasawholehavenot
beenestablished,northoseofinteractingneurons(althoughprogresshasbeen
andisbeingmade).Still,thereremainseverallevelsofabstractionbetweenthe
activityofneuronsandafunctioningdecisionmakingautonomousmind.And
15
approachesthatattempttomodeltheintermediatestageofcognitive
architecturehavemuchthesameproblem.Theyraisequestionsaboutifany
modelaccuratelyrepresentstheprocessoccurring,asopposedtosimply
replicatingthepatternofactivation.
Itmightbeimaginedthattheappropriatereactionatthispointwouldbethatno
approachcanaccuratelystudythephenomena,butthiswouldbetomissthe
largerpicture.Alloftheapproachesdetailedabovehaveprovidedandcontinue
toprovidevaluableinsightintotheoverallpicture.Top‐downapproachesbegin
tomaptheoutcomespaceandprovideinsightforcognitivearchitecture.Inturn,
thissuggestspossibleactivationpatternstostudyforinneuroscience,thattests
thesehypotheses,andfeedsbackintotheotherapproachesaswellasinspiring
newideasforautonomyandsoforth.Noneoftheapproachesalonecanaccount
forsuchacomplexidea,buttogetherthedisparatesourcesofinformationcan
starttobuildupanaccuratepicture.Thisthesis,beinginterestedintheroleof
contextintheprocesswillcomefromatop‐downperspectiveandseekto
providesomeinsightfromthatpositionasitdoesso.
Structure
Thisthesiswilladdresstheproblemofhowcontextaffectsdecisionmaking,
particularlyinformationpresentationinongoingtaskscenarios.Thischapterhas
hopefullyprovidedsomegeneralbackgroundandrationalefortheideasthatthe
restoftheworkisnowgoingtoexploreinamorescientificmanner.
ChapterTwoWillprovideamorein‐depthreviewoftherelevantliteraturethat
thisworkhasdrawnupon,andstarttoboundthegeneralproblem,identifying
whereempiricalworkcanusefullystarttoaddressthisproblem.Inparticular,
theareaofboundedrationalityandframingeffectswillbeconsideredasa
mediumforexamininghowcontextaffectsdecisionmakingintasks.Itwillbe
suggestedthatitisnecessarytoredefineframingtoaddressthisproblem,anda
roughsetofparameterstothisendwillbegiven.
ChapterThreewillthencoveraninitialexploratorystudyutilisingamodified
versionofthecardgameSolitaireasitsexperimentalmediumtoseehow
16
informationpresentationcanaffectdecisionmakingandstrategicchoices.
Generalprinciplesofthemethodologicalapproachbeingusedarealsoexplained,
andaninitialmodelofadaptationisposited.
Chapterfourthenutilisesthisunderstandinginthefirstfullexperiment.It
againadoptsacard‐gamescenario(althoughadifferentone,basedonthecard
game‘Uno’)asitsmediumandusesthistoexaminebothdecision‐making
adaptationinanongoingtask,andalsohowboundedrationalityworksinthis
novelconstruct‐thatitdoes,andthatitalsopossessessomeuniqueproperties.
Havingidentifiedthatthereareuniquepropertiestoboundedrationalityinan
ongoingtaskthatarereflectedinthewaythatdecisionsaremade,ChapterFive
appliesadifferentmethodologyandframingmechanismtothesameconcept,
nowutilisinganarrativeprobleminvolvinganoutbreakofavirulentdiseaseata
musicfestival.Itintroducestheconceptof‘confidence’indecisionmaking,and
usesittoprovideanadditionaldimensionforanalysis.Discoveringthat
confidenceisaffectedbyframing,andthatframe‐positionisalsoimportant,
ChapterSixre‐appliesthisunderstandinginanothernarrativetask,thistimein
thedomainofcyberinfluence.Furtherdiscoveriesaremadeaboutthe
persistenceofframingeffects,bothintermsofthedecisionsmade,andhow
participantconfidenceisaffected.
Finally,ChapterSevensummarisesthefindingsfromthroughoutthethesis.The
modelofadaptationisreconsidered,andinparticular,thestagesofformulation
andintegrationarere‐describedinmoredetailbasedonthediscoveriesmade.A
theoryofframingintasksislaidoutandaredefinitionofframingisspecified
basedontheevidencepresentedinpreviouschapters.Theimplicationsforhow
wemakedecisionsareconsidered,andsomeapplicationsandfurtherwork
consideredbeforefinallyconcluding.
17
CHAPTERTWO‐Literaturereview
Theprocessofmakingadecisionaccordingtologicalprinciplescanbeexplained
andrenderedpredictableandreplicablewithrelativeease.However,atthesame
time,suchawayofmakingachoicecanbeslow,over‐cautiousandinefficientin
therealworld,anddoesnotreflecthowmostaremade(Evans,2005).Theother
sideofdecisionmaking‐quickanddirtymethods,assumptions,heuristicsand
guesses–hasbeenextensivelystudied,butthereremainmoreunanswered
questionsabouthowitworks.Thisisthesideofdecisionmakingthatthisthesis
willbeaddressing.Thischapterwillbrieflyconsiderissuesofcategorizationand
biasinhumanmentalprocessesasabackgroundtoheuristicreasoning.Itwill
thenexplorepreviousresearchintostrategicproblemsolvingbeforeexamining
thetheoryofboundedrationalityandframingeffectsspecifically.Anapproach
willthenbeoutlinedtoutilizeframingeffectstoexplorehowcontextinforms
decisionmakinginongoingtasks.
Categorization
Categorizationistheprocessbywhichmentalrepresentationsdetermine
whetheranentityisamemberofacategory(Medin&Rips,2005).Itisalsoone
ofthefundamentalprocessesbywhichpeopleunderstandtheworld.Wegrasp
conceptsthroughunderstandinghowonethingislikeanother.Thisprovidesthe
foundationabilitiessuchasmakingreasonableassumptionsbasedonlimited
evidenceorheuristicstodetermineabestpathforwards(Vosniadou&Ortony,
1989)byestablishingtheabilitytodeterminewhensomethingis‘closeenough’
towork.Sinceinformationisfrequentlyincompleteorambiguous,best‐guess
processesrepresentamajorityofthedecisionswemake(Hogarth,1987).
Exactdefinitionsandusesvarybyresearcheranddiscipline,butitisgenerally
agreedthatcategoriesaredefinedbyconcepts,andconceptsaremental
representationsusedtodistinguishbetweencategories.Sothecategoryof
‘squares’isdefinedbytheconceptof‘havingfoursides’(Rosch,2004).Thiscan
besomethingofarecursiveeffect,wheretheconceptofonecategoryisa
categoryofitsownandso‐on.Indeed,intheexamplegiventhereisarelianceon
18
thecategoryof‘side’havingalreadybeenestablishedtodefine‘square’.The
difficultyofpreciselydefiningtheseideas(andtheirinterdependent,somewhat
recursivenature)hasbeenasourceoffrustrationtoresearchersformanyyears.
Theirprecisenatureandboundarieshaveprovedelusivetolockdown,evenas
theunderlyingideasappearfundamentaltohumancognition(Lakoff,2004).The
representationisintuitivelyappealing,andhasbeenusedasthebasisfora
numberofmodelsofknowledgestructuressuchasthehierarchical
representationproposedbyCollins&Quillian(1969).Thismodelincorporated
elementsofnetworkmodelling,butusedcategoriesandconceptsassomeofthe
basicbuildingunits.Itwasfoundtobeaninsufficientmodeltoexplainhuman
memoryoverall–negativerelationshipsareidentifiedfasterthanitwould
predict,andtherearereversalsofthepredictedcategoryeffectinsome
instances(dogsaremoreeasilynamedas‘animals’than‘mammals’forinstance)
(Rosch&Mervis,1975)–butitprovidedastartingpointthathassincebeen
extended.Itestablishedthatwhilstincomplete,associationsareausefulpartof
modelinghumancognition.
AssumptionsandBiasing
Oneofthenaturalconsequencesofcategorizationbeingpartofthebasisfor
humanmemoryisthatattributesthatareassociatedwithagroupareappliedto
individuals–whichcanleadtoincorrectassumptionsandbias.Agenerally
appropriateconceptforthecategory‘birds’isthatthey‘canfly’–though
penguinsandostrichesdonotfollowthisrule.Thiscanleadtoassumptions
whenmakingdecisions.Forexample,whenataneventsuchasafestivalorrace
daymostpeoplewillassumethatsomeonewearingfluorescentclothingisan
officialthattheycanaskforinformationsuchasdirections.Thisisnot
necessarilytrue‐anyonecanbuysuchclothingiftheywishto,andtherehave
beenexamplesofwherethishasbeenemployedinordertofoolpeopleand
commitfraud.Itisanexampleofhowcontextandover‐categorizationcanlead
toincorrectassumptions,butalsohowsuchassumptionsaregenerallyaccurate,
andthuswhyweusethem.
19
Asexperimentaldemonstrationofthistypeoftendency,researcherssentequally
qualifiedCVswithdifferentnamesattachedtothemtoalargenumberof
companiesintheUS.TheyfoundthatregardlessofwhichCVthenameswere
attachedto,applicantswith‘black‐sounding’nameswerelesslikelytobeoffered
aninterviewthanthosewith‘white‐sounding’names(Bertrand&Mullainathan,
2004).Similarly,amongmentalhealthprofessionalswhereitisexpectedthat
diagnosiswouldbepurelyobjectiveandevidence‐based,therehasbeen
evidenceofclass‐baseddiscriminationwithlower‐classpatientsbeing
prescribedmoreseverepsychiatricdisordersthanmiddle‐classcounterparts
(Routh&King,1972).Thefactthatthisdifferenceispossibleisatleastinpart
duetothefactthatpsychiatricdisordersaremorereliantuponthejudgmentof
thepractitioner,butitstillillustrateshowanimplicitbiascanaffectdecision
making.Inbothcasesthepeoplemakingthedecisionsinthestudiesrejectedthe
ideathattheirdecisionswerebeingmadeonanythingotherthanthefacts
provided,buttheactualresultsillustratedanunderlyingbias.Thiscanbe
explainedasaconsequenceofhowpeoplemanageknowledgeandaresultof
categorization,andalsoillustrateswhycontextcanbeimportanttotheoutcome
ofdecisionmakingevenwhentheinformationmightseemextraneousor
unrelated.
Biascanalsoderivefromheuristicreasoning.Oncesuchexampleisconfirmation
bias‐thetendencytoseektoconfirmratherthandisproveabelief.Whenpeople
withstrongopinionsonthedeathpenaltyweregiventhesamefabricated
studiestorateforaccuracyandinformation,theyratedtheinformationthat
confirmedtheirpriorbeliefasmoreconvincing(Lord,Ross,&Lepper,1979)
despitethefactthatalltheinformationwasuntrue.Amoredirectexampleof
thisistheWasoncardtask(Wason,1966),inwhichparticipantshavetotesta
ruletoseeifitcorrectlydescribesasetofcardswhichhavenumbersononeside
andlettersontheother.Logicallyparticipantsshouldseektoattemptto
disprovetherule,butinsteadtheytendtopositivelyconfirmit–whichwillnot
actuallyinformthemiftheruleistrueornot.Aninterestingfollowuptothis
finding,however,isthatparticipantsincreasetheirabilitytofindthecorrect
20
answer(uptoabout80%fromaround20%)iftheproblemiscoachedinreal
worldterms.Againcontextcanbeseentoaffecthowthesameproblemissolved.
StrategyandPlanning
Thatcontextcanaffectdecisionmakingandtheprocessingofinformationis
clearfromabove,butraisestheotherhalfofthequestion;whataboutstrategy?
Inthecontextofthiswork,strategywillbetakensimplytomeanplanningand
decisionmakingforataskthatgoesbeyondasingledecisionpoint.Thisareais
lesswelldefinedandempiricallyaccountedfor.
Strategyhasbeendescribedasafunctionofstimulusdetectability,whereoncea
stimulusisdetected(generallyagainstabackgroundofexperimentalnoise)it
promptsaparticularstrategicresponseintheformofaparticularschemabeing
enacted(Tanner&Swets,1954).Thishasbeensupportedforavarietyofsimple
cognitivetasksaswellassomemorecomplexonessuchaschoicereaction
(Sperling&Dosher,1986)andprovidesadecentaccountofaction‐levelstrategic
activity.Itdoesdownplaytheinternaldeliberativeprocesshowever,andrelies
heavilyupontimingresponsestostimuliforevidence‐measuresthatbecome
lessreliableorimportantwheremorecomplexdecisionsarenecessary.Forthis
reasonthisthesiswillbeconcernedwithhigher‐levelproblems.
Extensiveusehasbeenmadeofgamesinstudyingongoingproblemsolving,
althoughthesehavetheirownlimitations.Generallytheseproblemsare
‘solvable’,whereaparticularstrategycanbequantativelyestablishedasoptimal.
ConnectFourwassolvedaspartofamastersthesisforinstance(Allis,1988).
Similarly,whilstchesshasthusfardefiedanysuchfullaccount,thebesthuman
competitorshavebeenbeatenbyartificialcounterpartswhodrawprimarilyon
advancedsearchprinciplessuchastheSUPREMarchitecture(Berliner&
Ebeling,1989).Impressiveasthesedevelopmentshavebeen,however,they
utilisesearchtrees,andexploitthefactthattheenvironmenttheyareworkingin
islimited,definableandthereforecanbeaccuratelymodeledwithlittle
ambiguity.Solvableproblemsdonotnecessarilyrepresentthesortofproblem
encounteredintherealworld,northemannerinwhichahumanapproaches
them.Thisthesisisconcernedwithsituationswherethereeitherisno
21
solvability,oritisimpracticaltodetermine,andthestrategiesthatexistin
humancognitiontocopewiththat.Thereisevidencethatpeopledoshowsome
ofthesesamecharacteristicsasthesolutionsproposedbysolveableproblem
models,withincreasingskillassociatedwithincreasingsearchdepthinhuman
players(Gobet,1997).Howeverwhilstthiscanaccountforaspectsofhuman
problemsolving,itfailstocapturethewholeprocess.Astudycomparing
differenttypesofinstruction(conceptualversesprocedural)forchessnovices
didfindthatitwasconceptualknowledgethatleftparticipantsbetterprepared
tofindsolutionsformateproblems(Marmeche&Diderjean,2001),providing
someevidencethatitistheabilitytounderstandaproblemratherthanjust
memorizethenecessarycomponents,thatallowsadaptationtoanewproblem.If
optimizedsearchmethodologieswereallthatwasrequired,procedural
informationshouldhavealsoprovedequallyuseful.
Intryingtorepresenthumandecisionmaking,acommonapproachhasbeento
definetheprocessasaseriesofparts,particularlyinmanagementstudies
(Kepner&Tregoe,1965)–analysis,identification,causes,objectivesandso
forth.Buttheseapproachesarelimitedbythefactthattheyaregenerally
attemptingtoreplicate‘correct’decisionmakingandprovideaprescriptionfor
otherpeopletofollowratherthananaccountofcognitivearchitecture.More
generally,whenproblemsolving,peopleareseentoadoptbasicstrategiesor
planswhenapproachingaproblemthatarethenmodifiableoncontactwithnew
information.ThisformulationwaspartofthebasisforACT‐R,anaction‐level
accountofcognitivearchitecture(Anderson,Matessa,&Lebiere,1997)which
hassincebeenextendedandbuiltuponextensively.Adoptinganewrule
suppliedbyahintcanbeobservedandmodeled(H.A.Simon&Reed,1975),as
cantheeasingofdifficultyonceaparticularskillorruleisinternalized
(Kotovsky,1985).Butevenhere,theimportanceofcontextisstillpresent.
Domainspecificityisawellestablishedphenomenawherethesolutionthatis
obtainedcanrelyheavilyonthecontextinwhichtheproblemisposed
(Hirschfeld&Gelman,1994).Thetheoreticallysameabstractproblemcanbe
eithereasilysolvedoralmostimpossibledependingonwhetherpeoplehave
experiencedthatparticularcontextbefore.Thisalsoextendstocreativity,with
22
theworkproducedvaryinginqualityandquantityaccordingtothedomainin
whichaquestionisposed(Reiter‐Palmon,Illies,Cross,Buboltz,&Nimps,2009)
inasetupwherethereareno‘correct’answers.Formalmodelsofproblem
solvingsuchasACT‐Rdonotprovideaccountsofthisproperty.
SituatedCognition&ProblemIsomorphs
Situatedcognitionisdefinedasdifferencesinresponsethatcanbeobserveddue
toenvironment(Kirsh,2009).Itdoesnotattempttomodeltheinternal
processesofaproblemsolvingapproach,butratherthathowaproblemis
solved(andhowa‘problem’canbedefined)isafunctionofthecontextinwhich
itisfound–physical,socialandmore.Itarguesthatthementalapproachto
situationscannotberepresentedsimplybyrelatingthembacktoabstract
principlesbutaregroundedinthesituationtheyarefound.
AfamousexampleofthisconceptisastudyonBrazillianstreetchildren,who
haddevelopedparticulardomainspecificproceduresformathematical
functions,andfoundithardtotransferthisunderstandingtomoreformalized
mathematicalsettings(Carraher,Schliemann,&Carraher,1985).Asimilar
examplefoundmilkmenmadeuseofthephysicalshapeofthecontainersin
whichtheydeliveredmilktodocalculationsandagainfounditdifficulttodo
equivalentlydifficultcalculationsoutofthiscontext(Scribner,1984).Bothof
theseexamplesdemonstratehowproblemsolvingisnotnecessarilyreducedtoa
consistentabstractrepresentationthatisthenre‐appliedasappropriate.Ifit
were,bothexamplesetsofpeoplewouldhavefounditaseasytodocalculations
outoftheirfamiliarcontextastheydidwithin.
Thedegreetowhichcognitionissituatedorinternalandabstractisamatterof
continuingresearch,buttheideathatpeople’smentalprocessesareaffectedby
theworldaroundthemisbroadlysupportedbyavarietyofresearch.Human
ComputerInteractionasadisciplineandusabilityspecificallyhasdemonstrated
thatperformancecanbeaffecteddependingonhowdisplaysareorganized,for
instanceinambulancedispatchdisplays(Moore,Hayes,&Wong,2013)orhow
mobilephoneinterfacesaffectusersatisfactionandefficiency(Kim,Proctor,&
Salvendy,2012).Theconceptofaffordabilityandusabilityrestontheideathat
23
certaincontextssupportefficienthumancognitionmorethanothers(Norman,
2002).
Problemisomorphsareanexampleofproblemsthattouchonbothstrategic
problemsolving,andsituatedcognition.Problemisomorphsaresuperficially
differentrepresentationsoflogicallyequivalentproblems(Kirsh,2009).
However,thesesuperficialdifferenceshavebeenfoundtoaffectpeople’sability
toreachasolution.TheTowerofHanoiisalogicproblemthatinvolves
attemptingtomoveasetofitemsfromonelocationtoanotherwhilstabidingby
rulesthatgovernhowthoseitemscanbeassembledinagivenlocation.Inthe
classicexampleitconsistsofastackofthreeitemsthatneedtobemovedfrom
onelocationtoanother.Onlyoneitemcanbemovedatatimeandthereisonly
oneintermediatespaceavailable.Ithasbeenfoundthatdifferentisomorphsof
thisproblem(wheretheitemsarenotnecessarilyorderedbysizebutbyother
propertiesfollowingthesamerulessuchasnumberorcolourinstead)take
significantlydifferentnumbersofmovestosolve(Kotovsky,1985).Thisshows
contextintheformofproblemrepresentationaffectinghowaproblemisboth
perceivedandapproachedeventhoughtheabstractlogicaltaskbeingperformed
remainsthesame.Similarlydifferentinteractionconstraintsonthesamelogical
problemhavebeenfoundtoaffectproblemsolvingalthoughthelogical
informationavailablethroughthoseinteractionsremainedconstant(Dou,2010).
Representationsofthesamepuzzleindifferentdomainshavefoundtoaffect
problemsolvinglikelihoodsignificantly,withphysicalanddigital
representationsaffectingperformance,andevendifferentdigital
representations(Kotovsky&Simon,1990).Again,thispointstosituated
cognitionandcontexthavinganeffect,aseventhedifferencebetweenbeingable
toholdsomethinginyourhandsormakingchangestothesameproblem
representedonascreenmatters.Theeffectsofisomorphismarealsonotlimited
todifferencesinhowataskisrepresented.Instructionsthatconveyequivalent
informationhavealsobeenshowntocreateperformancedifferencesindifferent
tasksonthesametool(Bibby&Payne,1993).Differentgroupswereshownto
performbetterindifferenttasksdependingonwhichinstructionstheyreceived,
24
indicatingthatinstructionscanaffecthowknowledgeisrepresentedinthemind,
andthushoweasilycertaintasksareperformed.
Thisworkisnotwithoutitslimitshowever.Workinthisareatodatehasagain
reliedupon‘solveable’problemstocreateisomorphicequivalents.Theseare
taskscenarioswithdefinableendsstatesandlimitedactionpossibilities
specificallychosensothatequivalencescanbecreatedandmanipulated.In
manywaysthisfocusissimplyreflectiveofarequirementofisomorphism,and
problemscannotbeshowntobeisomorphicwithoutbeingcontrolledinsucha
manner.Butmanyrealworldproblemsarenotsolveableinthiswayanddonot
necessarilyevenhaveanendstate,orparsableprocessesandstagestobe
quantified.Thesecanstillinvolvestrategicchoicesanddifferentrepresentations,
butcannotbesoeasilycontrolledandbalanced.
Whattheliteratureonproblemisomorphsillustrates,however,isthatlogical
approachesandstrategicchoicescanbeaffectedbythecontextinwhichtheyare
presented.Itthereforeseemslikelythatthisistrueforproblemsthatcannotbe
quantifiedinsuchamanneraswell,althoughamethodologyisrequiredwith
whichtoapproachthem.
BoundedRationality
BoundedRationalitywasfirstproposedbyHerbertSimon(H.Simon,1957),and
proposesthatratherthanbeinginherentlyrationalactors(asmanymodelsof
humanbehaviouranddecisionmakinginavarietyofdisciplineshavedone)
peopleactrationallywithintheboundsoftheircomprehensionoftheworld–
boundssuchascognitivebiases,perceptuallimitationsandcomprehension
failures.Boundedrationalityisnotconcernedwithhowinformationitself
persuadesadecisionmakertoonesetofactionsoranother,butratherhowthe
boundsthatthatparticipantareinareresponsibleforthevariabilitythatcanbe
observed(Gigerenzer,2008).Whatthismeansisthatthereisabodyofworkthat
existsalreadyestablishingdifferentconditionswhichareknowntoaffect
particulardecisionswithoutalteringtheinformativecontentavailableto
decisionmakers.
25
SystemsOneandTwo
Boundedrationalityisgenerallyseentooperateaspartofadual‐systemmodel
ofhumandecisionmaking.Thesearereferredtoassystemoneandsystemtwo,
andreflectdifferentaspectsofthedecisionmakingprocess,adistinction
originallycoinedbyStanovichandWest(Stanovich&West,2000).Theseare
summarizedbyKahneman(Kahneman,2002)as:
Systemoneisthequick,instinctivedecisionmakingsystem.Itisdescribedas
fast,automatic,effortless,associativeandslow‐learning.
Systemtwoisthetimeconsuming,consciousdecisionmakingsystem.Itis
describedasslow,serial,controlled,effortful,rule‐governedandflexible.
Thesesystemsarenotaneither/ormodel.Whilstsomequestionsmayuseone
morethantheother,theyarethoughttoconstantlyinteract,andoneoftheroles
ofsystemtwoistoprovideexplicitmonitoringofsystemone(Gilbert,2002).In
differentsituationsonesystemwilltakepriorityovertheothereventhoughthe
otherisstillactive.Forexample,ifapersonisaskedwhat1138+435issystem
twowilltakepriorityaswouldbeexpected.However,ifaskedtodecidebetween
1600and2300whichisclosertothecorrectanswerforthesamequestion,
systemonewillallowmostpeopletoanswer1600asit‘feels’morecorrect.
Thesesystemsarethebasisfortheboundsofboundedrationality.Theydescribe
howthefirstsysteminparticularisaffectedbytheconstraintsthatsurroundit,
andthenalsohowthetwosystemswiththeirdifferentmethodologiesand
strengthsinteract–when,whereandhow.
Thishypothesishasbeenputtoextensiveusewithinthefieldofdecision
making.Ithasbeenusedtoprovideinsightformarketing,explainingpatternsof
buyingonthestockmarketandpersuasioninpolitics(Myers,2002).Therehas
alsobeensomesupportfromitwithinneuroimagingscans,whereparticular
activationpatternshavebeenfoundforgutfeelingsandinsight(Bowden,Jung‐
Beeman,Fleck,&Kounios,2005).Otherworkhasfoundactivationinthemedian
orbito‐frontalcortex(whichreceivesinputfromalloverthebrainandthusisa
plausiblelocationforassemblingdisparatesources)andtheamygdala(whichis
26
thecenterofemotionsinthebrainandthereforerelatedtotheemotional
componentofsystemone)instudiesofinstinctiveguessing,providinga
neurologicalbasisforthedistinctsystems(Volz&Cramon,2006).Therefore,the
two‐systemhypothesishasareasonablegroundinginbothpracticalapplication
andintheoreticalneurosciencestructures,whichinturnraisesthequestionof
howboundedrationalitycanhelptoexplainhumanbehaviourandfurtherthis
work.
ProspectandUtilitytheory
Asanexampleofhowboundedrationalitycanexplaindecisionmakingwecan
usetheexampleofprospecttheoryandhowithasextended(andtosomedegree
supplanted)utilitytheory.
Utilitytheoryisatheorywhichmodelshowchoicesshouldbemade,basedona
setoflogicalaxioms(Myers,2002).Ithasprovidedthebasisforalotof
economictheory,asitisassumedtonotonlydetailhowdecisionsshouldbe
madebutalsohowtheyaremade.Itassumesthatthereismathematicalparity
betweentherelativefallsandgainsinwealthandthatwealth’sperceivedutility
orvalue,aswellasthevalueofvariousobjects.Ifapersonisofferedachoice
betweenacupofcoffeeandacupofteaandshowsnopreference,forinstance,
thentheywouldalsohavenopreferencebetweenthechoiceofa40%chanceto
winacupofcoffee,anda40%chancetowinacupoftea.Iftheylikedthetasteof
coffee,butnotteathentheywouldsimilarlypreferthatchanceovertheother.
However,asweknow,peopledonotalwaysmakedecisionsbasedonstrict
logicalprobability.Considerabetwhereacoinistossed.Achoiceisoffered
betweeneitherwinning£100ifthecoincomesuptailsortaking£46forcertain
(Kahneman,2011).Mostpeoplewilltakethesurething,despitethefactthatona
statisticallevelthechanceisactuallymorerewarding.Consideralsoifthisbet
wasofferedatthelevelof£1ifthecoincomesuptailsor46pforsure–more
peoplewouldnowbewillingtogoforthechance.Finally,imagineifthesame
conditionswereoffered,butthelevelswerenow£10,000or£4,600–evenfewer
peoplewouldnowbelikelytotakethebetratherthanthesurething.
27
Theexplanationgivenforthisinutilitytheoryisoneofpsychologicalintensity.
Ratherthanbeingcalculatedpurelyonthemeritofstatisticallikelihood,people
evaluatethesebetsbasedontheintensityoftheoptionsweightedbytheir
probability.Whatthismeansmoresimplyisthatthedifferencebetween46pand
£1isnotactuallyverymuchbecauseneitheristhatvaluableandneitheristhat
differentfromthebaselinestateofzero.Hence,thechanceisnotseenasthat
important.However,whenthestakesareraisedto£10,000and£4,600suddenly
thepsychologicalintensityofthestakesaremuchhigherandthechanceismuch
moreofarisk,sopeoplewilldefertothesurethingwithincreasinglikelihood.
Thisisthecoreofutilitytheory;thevalueofabetorpropositioncanbe
objectivelycalculatedaccordingtothevalueswithin.
Thistheorywas,andtoanextentis,stillthebasisforeconomictheoriesofvalue
formanyyears,butitlacksabasicfactorthatwouldaltermanyofitspredictions
–referencepoints.ThisflawwasfirstidentifiedbyKahnemannandTversky,
whopointedoutthatitisnotsimplytheutilityofwealththatdrivessuch
decisions,butalsowhereapersonstarts(Kahneman&Tversky,1973).Itisthis
observationwhichformsthebasisofprospecttheory.
Toillustratethis,consideranotherbet:Winning£2millionguaranteed,ora50‐
50chancetowineither4millionor1million.Theutilityofthesetwo
propositionsisequal,anditwouldnotbesurprisingtoseeapersonpickeither
oneortheother–theincreaseinutilitybetween1millionand4millionisnot
actuallythatmuch.
Nevertheless,nowconsiderthetwopeopletowhomthisbetisactuallyoffered.
Jackhas1milliontostartwith,andJillhas4million–andinparticipatinginthis
bettheyhavetostaketheircurrentwealth.Theutilityofthepropositions
remainsexactlythesameasbefore,butthecontexthaschangedmassively.For
Jackthebetisessentiallyano‐losescenario–hecannotexitwithlessthanhe
startedoffwith,andwhilstbothpropositionscanofferhimmore,thechance
offersmore.Intheworstcaseheiswherehestarted.Thismakeshimmorelikely
toriskthechance.Heisprimedtoengageinrisk‐seekingbehaviour.
28
Jillontheotherhandismostlycertaintolosesomething.Atbestshecould
preservewhatshehas,andinthatcaseshewouldbeaslikelytogodownto1
million.Fromherperspective,sheislookingtominimizeherpotentiallossesand
isthereforemorelikelytogoforthe2millionguaranteed.Sheisprimedtobe
risk‐adverse.
Thesepositionsaresupportedbyempiricaltestingputtingparticipantsinthose
virtualroles.Whatisimportantaboutthisresultisthatthestartingpositionof
thetwoparticipantschangestheirapproachtotheproblem,despitethefactthat
theirpotentialoutcomesareidentical(Kahneman,2011).Theirpriorpositions
shouldnotmatterduetotheutilityoftheoptions,andlogicallytheydonot,but
whenpeopleareinvolvedthecalculationsdochange.Contextaffectschoice
whenthechoicesthemselvesareperfectlybalanced.Thisisthecoreofprospect
theory–thatitmatterswhereapersonstartsfrom.
Anotherexampleofthisprincipleistoconsideranotherpairofpeople,George
andVicky.Georgehas8millionpounds,andVickyhas2million;bothinvested
onthestockmarket.Asaresultofavolatileday,VickyandGeorgenowboth
have5million.Whoishappier?NaturallywewouldsayVicky,asherwealthhas
increasedbythreemillion,whereasGeorge’shasdecreasedbythesameamount,
buttheiractualvalueisequal–byutilitytheorytheyshouldhavethesame
satisfaction.Whatisoccurringhereisthattheyareexhibitingriskaversion,
wherepeoplearemoresensitivetolosingwhattheyalreadypossessthanthey
aretogainingadditionalthings(Novemsky&Kahneman,2005).AgainUtility
theorywouldtreattheseasequalpropositions,butprospecttheoryrecognizes
theimportanceofthestartingpositiontothepsychologicalevaluation
undertakenandbetterpredictsbehaviour.Itisnotnecessarilyineffectallthe
time:ifinstructedto‘actlikeamarkettrader’participantsshowlessloss
aversionwhichsuggestsitisafunctionofsystemoneandcanbeselectively
applied(Sokol‐Hessneretal.,2009).Someevidencehasbeenfoundforthis
tendencytobeneurologicallybased,withabroadsetofgain‐sensitiveareas
(includingmidbraindopaminergicregionsassociatedwithrewardmechanisms)
showinglessactivationwhenpotentiallosseswerecontemplated(Tom,Fox,
Trepel,&Poldrack,2007).
29
Thekeyrelevanceofthisexampleforthisthesisisthatthestartingpointmatters
totheperceptionofgainorloss.Contextwillcolouraperson’sperceptionofa
gain/lossevent,andthuspresumablyaffecttheirreasoningwhenconsidering
whattodonext.
FramingEffects
Aparticularlyusefulcomponentofboundedrationalityforthepurposesofthis
workisframingeffects.Framingeffectsarewherethesameuseful,informative
contentispresentedindifferentwaysinordertoinduceausertowardsa
particularchoice(Gigerenzer&Selten,2000).
ConsiderauniversityEngineeringdepartmentwhoseintakeofundergraduates
wassplitinto20girlsand48boys,withtheyearhavingjustpassed.Theirnew
intakethiscomingyearwillbe30girlsand45boys.Thiscouldbeexpressedas
‘TheproportionofgirlstakingEngineeringhasincreased’.Alternatelywecould
say‘TheproportionofboystakingEngineeringhasdeclined’.Bothstatements
wouldbetrue,andbothwouldbelogicallythesamething,containingthesame
informationabouttheworld,buttheyfeelverydifferentstatementswhenread.
Tohearaboutmoregirlstakingengineeringmightsummonupsatisfactionthata
long‐presentgendergapwasapparentlydeclining.Tohearthatfewerboyswere
takingthesubjectmightraiseconcernsthattheyarefallingbehindtheirmore
studiousfemininecounterparts.
Thisisanexampleofaframe.Thesamelogicalinformationisimparted:aperson
whoreadseitherofthetwostatementswillhavethesamelevelofknowledge
aboutthestateoftheworldlogicallyspeaking,yetvastlydifferentperceptionsof
whatisgoingonintheworldofengineering.
PerhapsthemostfamousexampleofaframeisthatoftheAsianfluexperiment
(Tversky&Kahneman,1981).Inthisexample,participantsarepresentedwitha
choiceoftworesponsestoanoutbreakofAsianfluinapopulationof600people.
Theycaneitherguaranteethatacertainnumberofthepopulationwillsurvive
andacertainnumberwilldie,ortheycantakeaprobabilitythattheentire
populationwilleitherliveordie.Theseareexpressedas:
30
FormulationOne:
IfProgramAisadopted,200peoplewillbesaved.
IfProgramBisadopted,thereis1/3probabilitythat600peoplewillbe
saved,and2/3probabilitythatnopeoplewillbesaved.
FormulationTwo:
IfProgramAisadopted,400peoplewilldie
IfProgramBisadopted,thereis1/3probabilitythatnobodywilldie,and2/3
probabilitythat600peoplewilldie.
Notethatbetweenthetwoformulationstheactuallogicalinformationbeing
presentedisthesame.Alsonotethebetweenthetwoprogramsthestatistical
valueofbothisactuallyidentical–ifthesituationwasrun100timesandthe
sameprogramwaspickedeachtime,statisticallyspeakingthesamenumberof
peoplewouldbeexpectedtosurviveanddieineachchoice.
Itwasfoundthatinformulationone,72%ofparticipantschoseprogramA.In
formulationtwo,78%ofparticipantschoseformulationB.Itissuggestedthat
thereasonforthisisthattheframingaltersthetypeofdecisionbeingmade.
Formulationonesupportsarisk‐adverse(seeearlier)decision–participants
chosetobesureofsaving200people.Formulationtwoontheotherhand
supportsrisktaking–the400peopledyingforcertainisregardedasless
acceptablethana2/3chanceofeveryonedying.Theframeisthereforepriming
differentsortsofarisk‐takingheuristicthroughthewaytheinformationis
structured.Context,again,isdrivingthemannerinwhichadecisionismade.
Framinghassincebeenappliedtoarangeofcircumstancestoeitherexplain
decisionmaking,ortoillustratethattheprocessisnotasobjectiveaswemight
imagine.Ithasbeendemonstratedusingmortalityratestoaffectthedecisionsof
doctors(McNeil,Pauker,Sox,&Tversky,1982)andrichnessofadescriptionto
affectacourtdecision(Shafir,1993)forexample.
Theeffectsofframingarenotimmutableandunmovedbyotherfactors
however.Individualdifferencescanplayaroleforinstance.Framingmonetary
31
gainsorlossescannormallyprimeanindividualtoberiskadversewhena
choiceisframedasagain,butriskadversewhenthesamedecisionisframedas
aloss(Kahneman,2002).However,recentworkhasfoundthatwhilstolder
participantsdobecomeriskadversewhenastandardgain/lossgamblingtask
wasframedasagain,theydonotshowrisk‐seekingbehaviourwhenthesame
taskisframedasaloss(Mikels&Reed,2009).Similarlythereissomeevidence
thatexperiencecanpreventaframeaffectingadecision.Invitationstoregister
foraconferenceweresenttoallpeoplewhosubmittedapaperforaneconomics
conference,butthedescriptionofincreasingfeesasthedeadlinenearedwere
framedaseitherapenaltyforbeinglate,orabonusforregisteringearly.When
dividedbyexperiencelevel,juniorexperimentaleconomistsregisteredin
greaternumberswhenthechangeinpriceswasdescribedasapenalty,whilst
theirseniorcounterpartsshowednodifferencebetweenthetwoconditions
(Gächter,Orzen,Renner,&Starmer,2009).Theexperimentisnotablefor
establishingthataframecanbeeffectiveinarealworldsetting,andalsofor
showingthatexperiencemaycausepeopletobelesssusceptibletoframing,
althoughitshouldbenotedthatthismayrepresentknowledgeofframes,given
thatthesubjectpoolwaseconomistswhowouldbeexpectedtohaveahighrate
ofknowledgeofthesubject.Regardlessthefactthatitwaseffectiveonthejunior
subset(whoshouldalsohavethisdomainknowledge)suggeststhatabroader
effectmaybeatplay.Inalaboratorybasedexperiment,differencesinchoices
madewerefoundtoexistbetweensexes.Althoughbothsexeswereframed,they
respondedatdifferentratestoframeddecisionsindifferentdomainssuchas
moneyvstimeandlifevsdeathdomains(Huang&Wang,2010).Againthis
resultissomewhatambiguousforwherethereasonforthedifferencecomes,
anditisreasonabletosuggestthatsocialcueswhichsetdifferentprioritiesfor
thedifferentsexesmayberesponsibleratherthanfundamentalprocessing
differencesbetweenthesexes.Regardless,thesestudiesillustratethatpre‐
existingknowledgeandexperiencecanhaveanimpactonhowparticipants
respondtoframing,regardlessofthesourceofthatinformation.
Thereissomelimitedevidencethatframingcanoccurinanongoingtasksas
wellassingledecisionpoint,andthenbeaffectedbyadditionalfactors.These
32
experimentsdotendtohaveaverylimitedscopeforwhatcomprisesanongoing
task.Boundedrationalityhasbeenappliedintotheareaofstrategyselection,
particularlywiththesuggestionofcognitivelyboundedrationalanalysis(Howes,
Lewis,&Vera,2009).Thisprovidesanaccountofcognitivearchitecturein
psychologicalrefractoryperioddual‐taskperformanceandillustratesthe
potentialapplicationofboundedrationalityintocognitiveareas,butthat
strategyselectionaccountisatthecognitivelevelwherecontextisinvariant;this
thesisisconcernedwithactivitywhenthecontextinwhichastrategicdecisionis
beingmadevaries.Inanonlineshoppingtaskwhereoffersandpricereductions
wereframed,messagesthatwarnedagainstbiassuccessfullylessenedthe
framingeffectinhighlyengagedparticipants(Chenga&Wu,2010),althoughthe
effectwaslessenedbothbytheintensityofthemessagedecreasing,andbythe
participantbeinglessengaged.Regardless,asmall‐scaletaskwassuccessfully
framed,andthatframewassubsequentlycounteracted.Ina20‐daysimulated
stock‐buyingsimulationwhereparticipantsratedtheiremotionalstateasthe
taskwasconducteditwasfoundthataffectattenuatedframingeffects(Seo,
Goldfarb,&Barrett,2010).Thisbothagainsuccessfullyframedatasklonger
thanonedecision(albeitinahighlylimitedandartificialsetting)andlinked
framingeffectstoanothercognitivecharacteristic,inthiscaseemotion.Ina
complementarystudy,risk‐lossframedgamblingtaskparticipantswere
promptedtousecognitivereappraisaltechniquestoregulatetheiremotions
whilstperformingthetask,whichalsoreducedframing(Miu&Crişan,2011),
suggestingthattheemotionalcomponentmightnotjustbeattenuatingframing
butactuallydrivingit.
Thesignificanceofthesefindingstothisresearchisthatframingisacomplicated
andinvolvedprocess,onethatispotentiallynoteasilydivorcedfromthecontext
inwhichitoccurs,astheoriesofsituatedcognitionwouldsuggest.The
dimensionsofwhatcanaffectframing,andbeaffectbyframingappearto
potentiallybebroaderthansimplywhatdecisionismade.Framingtherefore
potentiallyrepresentsamethodologywithwhichtosystematicallyvaryhow
informationispresented,andthuspotentiallyaffectstrategicdecisions–with
thatdifferencethereforebeingattributabletopresentationratherthanlogical
33
contentoftheinformation.Researchhasdemonstratedthatframingcanexistin
tasksbeyondasingledecision,althoughasnotedthesetendtobeverytightly
controlledenvironmentsthatlackthecomplexitytosuggestthatstrategic
decisionsarebeingframed–thechoicesdonotgenerallyhaveongoing
consequencesforinstance.Thisoffersapotentialareainwhichtooffernovel
investigation.
LimitationsofFraming
Framingasatheoryhasnotbeenunchallengedhowever,andisnotwithoutits
conceptuallimitations.Theassumptionthatunderliesframingisthatthetwo
choicespresentedareequivalent(isomorphic)intheirambiguity,butthereare
suggestionsthatthismaynotbeentirelysupported.Areviewoftherelevant
literature(Kühberger&Tanner,2009)suggestedthattherelevantchoicesinthe
Asianfluexamplearenotequivalent,andthatwhenambiguityisequalized
betweenthetwodescriptions,theframingeffectiseithersignificantlyweakened,
ordisappears.Anexampleofthisisanexperimentthattransferredthedomain
oftheasianfluchoice,andcontrolledforthisimbalancedambiguityinthenew
context(Mandel,2001).Inthiscase,theframingeffectwasnolongerpresent.
Thesefindingraiseavalidpoint–whatexactlycanbeconsideredtobeaframe?
Iftheeffectsthathavebeennotedaresimplyacaseofunequalambiguitythen
canframingbesaidtoexistatall?Thestudiescitedabovefindfaultinthe
unequaldistributionofambiguity,butitisnotnecessarilythecasethattheir
definingitassuchisappropriate.Itispossibletocontrolforframingby
controllingtheamountofambiguitypresent,buttheprocessofdecidingwhatis
orisnotambiguousdependsitselfuponajudgmentcallbytheexperimenter,
andperceptionsofwhatisambiguouswillvarybetweenparticipants.Buteven
acceptingthatimbalancedambiguitymaybethesourceofthephenomenadoes
notinvalidateframingasanapproach.Realresultsandsignificantfindingswith
applicableimplicationscanandhavebeenfound.
Instead,framingmaynotbewellservedasbeingdefinedaslimitedtoprecise
andstrictlyisomorphicequalpairs.TheAsianfluexample(Tversky&
Kahneman,1981)wouldnolongerbeanexampleundersuchaclassification.If
34
suchrigidstandardsareappliedthenitcouldbecomefunctionallyimpossibleto
defineanythingasbeingframing–somelevelofunbalancecouldbefoundin
virtuallyanyattempttocreateapproximatelyequalsetsofinformation.
Suchastringentrequirementwouldmissthepointofframingonatleasttwo
levels.Firstlyframingisclearlynotsolelydefinedbyambiguity.Thestudies
presentedinthisreviewhaveidentifiedanumberofdimensionswhichcanaffect
thedecisionsmadeevengiventhesenon‐isomorphicsets.Sotheyremaina
viablemethodinwhichtoidentifyadditionalfactorsaffectingdecisionmaking.
Secondly,therealworldisacomplex,dynamicandlargelyuncontrollablesetof
conditions.Understandingrealworlddecisionsmayrequireacceptingadegree
ofimbalanceinmethodologies.Framingmaywellnotexistwithoutaslight
imbalanceofambiguity,butapproximateambiguitycancreatedemonstrable
resultsaslistedabove,andframingstillprovidesausefulmodelfor
understandingthoseproperties.Butthatsaid,itmaythereforebenecessaryto
re‐defineframingtoallowforuseinmorecomplicatedcontexts.
Re‐definingFraming
Therearestillunresolvedquestionstobeansweredabouthowtaskstrategyin
ongoingstrategictasksmightbeaffectedbyinformationpresentation.Bounded
rationalityprovidesevidencethatcontextaffectsdecisionmaking.Thereare
examplesofworkextendingframingintoongoingtasks,butthesehavebeen
tightlycontrolled,highlyartificialexampleswherethestrategicthinkingdoes
notreflectmorecomplex,ambiguousscenarios.
Framingeffectarestilllargelyconductedinasingle‐decisionparadigm,andas
partofthere‐definingsuggestedaboveitwouldbeusefultomoveitbeyondthis
paradigm.Kahnemanhasdescribedsuchchoicesasbeing‘thefruitfly’of
decisionresearch(Kahneman,2011),anditisanaptanalogyonanumberof
levels.Singledecisionsareeasilysetup,quicklyrunandrepeatable.Theyare
alsohighlycontrollableandfreeofmanyextraneousvariables–itwasa
relativelysimpletasktosequencethefruitflygenome.Theycantellusagreat
dealabouthowwemakedecisionsinthesamewaythatfruitfliescantellus
aboutgenetics,buttherearelimitspreciselybecauseoftheirsimplicity.
35
Inthesamewaythatfruitfliesareasimplegeneticblueprinttoworkwith,single
decisionsareasimpleparadigmtoemployandbothmissthelargerpicture.
Humansaresignificantlymorephenotypicallycomplexthanfruitflies,and
decisionmakingintherealworldissignificantlymorecomplicatedthansingle‐
decisiontasks.Butapplyingframesintoamorecomplicatedandecologically
validcontextisnotsimple.Framesrelyonpreciseconstructioninordertobe
consideredisomorphic,butsuchtightdefinitionslimittheirapplicabilityina
systemwherethosebasicconstraintsaremoreabletobechallenged,
contradictedorusurped.Approximate,ratherthanfullequivalencemaybe
necessarytoassessframinginthiscontext,butthatisnotaninappropriate
methodologytoapply–ashasbeennotedeventightlycontrolledsingle
decisionsmaynotbeequalintheirambiguitybuttheystillrepresentavalid
sourceofinvestigation,thiswouldbeanextensionofthat.
Thequestionhaspotentialimplicationsforwidertheoriesofboundedrationality
andnotjustframingeffects.Thepremiseofsystemsoneandtwois,as
previouslynoted,atleastpartiallythatsystemtwohasmonitoringdutiesover
systemone(Gilbert,2002).Soatwhatpointdoesthemonitoringsystemcutin
andoverridetheemotionalorgutdecisionthathasbeenmadeasatask
progresses?Someresearchhasaddressedtheconceptthatmonitoringcanoccur
(Gigerenzer,2008)butthishasgenerallytakentheformofprimingsubjectsto
respondtoquestionsandconsideringthemmoreconsciously.Alternativelyit
hasbeenshownthatintelligencecanbelinkedincertaintaskswithagreater
degreeofsystemtwocontrol(Frederick,2005),butagainthisisinasingle
decisiontask.Therehasbeenlittleattempttodeterminehowsystemtwomight
graduallytakebackcontrolfromsystemone.
Anobjectiveofthisthesiswillbetoprovideaframeworkforbetterdefining
frameswithinalargercontext.Asnoted,framingeffectsaregenerallyrelianton
tightlycontrolledsinglechoiceandlimitedcontextmethodologiesinordertobe
balanced,butsuchtightlycontrolledcircumstancesaretooartificialforthis
objective.Thereforeitwillseektore‐applymoretraditionalframesand
experimentalparadigmsascloselyaspossiblewhilstalsoexpandingthemto
morecloselyresemblecomplex,ambiguousongoingtasks,followingageneral
36
principleofobtainingasmuchapproximatebalanceaspossible.Fromthese
approximations,amorepreciseandquantifiedframeworkwillbedevelopedand
suggestedintheconcludingchapterofthisthesis.
37
CHAPTERTHREE–EquivalentRuleChanges:TheSolitaireStudy
Introduction
Thepreviouschapterestablishedtheexistingunderstandingofframingin
currentliterature,andillustratedthelimitationsofthecurrentdefinition.
Expandingthedefinition(andunderstanding)ofthisprocessisnottrivial
however‐thereexistagreatdealofconflicting,possiblyconfoundingfactors
withinpotentialexperimentaldesigns.
Thefirststudywasdesignedasbeingsolelyexploratoryinnature.Ataskwas
designedtobeapproximatelyisomorphicatthenewruleimplementationlevel
andforthefindingstoactasapointerforsubsequentresearch.Itwasnot
designedtobeperfectlybalancedforthetheoreticalreasonsdetailedinthe
previoussection;producingadirectisomorphofrulechangesinacomplex
environmentisextraordinarilycomplicatedandpotentiallyselfdefeating.A
generalhypothesiswasthereforeemployedratherthanspecificpredictions.It
washypothesizedthatapproximatelyisomorphicruleswouldproduce
observabledifferencesinbehaviour.
Thetaskwasdesignedtocontainanelementofnoveltyintheformofintroduced
rulealterations,butfamiliarityintheunderlyingmechanismsandmanipulations
beingemployed.Itwasalsodesignedtoallowadegreeoffreedomofbehaviour
andchoice,inorderthatalternativebehaviourswerepossible.
Design
Measurementswerekeptgeneralratherthantightlycontrolledorspecificby
design,aswhatshouldbeobservedwasnotyetestablishedandthehypothesis
wasgeneralratherthanspecific.
Thestudy’senvironmentwasrequiredtofitseveralrequirements.Itwas
intendedtobeafamiliartaskthatwasalsomodifiableinordertointroducethe
novelelement.Itwouldalsoneedtoberelativelycontrolled,witheasily
38
identifiedandcategorisedactionsandbehaviours.Finallyitwasdesiredthatit
wouldbeambiguous,atleasttothedegreethatoptimalbehaviourcouldnotbe
easilyandobviouslyadoptedinordertobeabletodifferentiatebetween
differentstrategicapproachestotheproblem.
Itwasdecidedtoutilisethecardgame‘solitaire’.Gamespresentacontrollable
andconstrainedenvironmenttostudybehaviour,wherethelimitsandfeatures
ofasystemcanbereadilyunderstoodandaccountedfor.Theparadigmalso
enablesfortheintroductionofanewruleasasourceofvariationtoafamiliar
environment.
Apotentialobjectiontothismethodologyisthatthegameiscomparatively
knowledge‐lightandsolvable(inthemannerthattheoriesofsituatedcognition
takeexceptiontotraditionallimited‐scopedecisionmakingexperiments(Kirsh,
2009)).However,thisisnotconsideredproblematicfortworeasons.Firstlythe
taskdomainstillrepresentsasignificantexpansionofpotentialaction.Secondly,
whilstsmallchangestocardgamescanpotentiallybemadeoptimalitdoesnot
followthattheycanbemadeoptimalandimplementedperfectlybyaperson‐
particularlynotwhilstplayingthegameforthefirsttime.Thechangesare
sufficientlycomplexthatthereisenoughambiguityformistakestobemadeand
imperfectandvariablestrategiesadopted.Additionally,theaimofthislineof
studyisnottoestablishabest‐fitsolutionforthisgame,cardgamesorindeed
gamesingeneral–butrathertounderstandthecognitiveprocessesinvolvedin
reasoningformorecomplexsituations.Themethodologyofgoingfrom
encounteringsomethingnewandthenendingupatasolutionarewhatis
interesting,nottheresultitself.Theprocessofdevelopingasolutionandtowhat
degreeitisprimedbycontextisstillrelevantwhenobtainedfroman
optimizablescenario,althoughtheeventualintentionistounderstanditinterms
ofnon‐perfectibleexamples.
Method
Participantswerepre‐selectedonthebasisofhavingfamiliaritywithsolitaire
simplybyaskingthemiftheyknewhowtoplayit.Otherwise,nocriteriawere
appliedtorecruitment,otherthanavailability.
39
Initially,allparticipantsfilledoutaquestionnairethatconsistedofaLikert‐scale
setofquestionsratingtheirconfidenceinandlevelofunderstandingofgamesin
generalandsolitaireinparticular(seeappendixone).Theythenansweredaset
ofopen‐endedquestionsaboutthegamedesignedtoelicitknowledgeabouthow
thegameworkedandparticularstrategies,aswellasgeneralunderstandingof
strategyandadaptationintheireverydaylife.
Solitaire
Solitaireisasingleplayercardgamethatisalsoknownaspatienceandfamiliar
toalargenumberofpeopleatleastinpartduetoitsinclusionasafreegamein
Microsoftwindowsoperatingsystemssinceatleastwindows3.11.
Thegameisusuallyplayedwithafulldeckofnormalcards(AcetoKingofall
foursuit),althoughforthepurposesofthisexperimentitwasplayedwithAceto
Queen,astheremovaloftheKingsmadethegamemarginallyeasiertoplayand
quickertocomplete.
Solitairehasasetofcoremechanics,althoughthereissomevarianceinhow
particularindividualsplay.Thereexist,forinstance,multipleoptionsforhow
manycardsaplayerdealsfromthedeck,andhowmanytimestheyareallowed
tocyclethroughsaiddeckinthecourseofthegame.Therulespresentedbelow
andusedinthisexperimentareonecommoninstanceoftheoptionsavailable.
Objective
Theobjectiveofagameofsolitaireistomovecardsfromtheplayingareainto
foursortedpilesofcardsascendingfromAcetoQueeninagivensuit.These
cardscanbemovedatanytimefromtheplayarea,butmustbemovedinorder,
startingwiththeacesandcontinuingupwards.Thecardscanbemovedback
downoncetheyhavebeenplacedinthewinningpilestobeusedinthegame
againaspartofalegalmove,butonlyiftheyarethetopcard.
Setup
Itwasnecessarytosetuptheplayingareacorrectlybeforestarting.Sevenpiles
ofcardsweredealtontothetableinalineinfrontoftheparticipant.Fromright
toleftthesepilescontainedanascendingnumberofcards,fromoneonthefar
40
righttosevenonthefarleft.Thetopcardofeachpilewasthenturnedoverso
thatitwasfaceupandleftontopofthepile.Thepileswerestaggeredslightlyso
thatitwasvisiblehowmanycardswereleftineachcaseundertheface‐upcard.
Abovethisarrangementofcards,fourspacesweredesignatedforcardstobe
movedintoaspartofthewinconditionofthegame.
Theremainingcardswereplacedfacedownonthetableinfrontofthe
participanttoserveasthedeckfromwhichadditionalcardswouldbedrawnin
thecourseofthegame.
Play
Playproceedswithaparticipantmakingalegalmove.Thereareseveralmoves
whicharelegal,andaplayercanperformanyoftheseanynumberoftimes
beforechoosingtodoanother.Thesemovetypesaredetailedbelow.
Deal–Aplayerdealsthreecardsinsequenceontothetable,orontopreviously
dealtcardsifitisnotthefirsttimethattheyhavedealt.Thesecardsaredealt
faceup.Aplayermaythenusethetopmostcardtomakealegalmove.Ifthey
succeedatthistheycanusethenextcarddown,andsoforth.Ifthereareless
thanthreecardsinthedeck,theydealasmanycardsareleftandplayfromthat.
Iftherearenocardslefttobeplayed,theplayerpicksupthepileofdealtcards
andturnsitoverwithoutshufflingandusesitasthedecktobedealtfromonce
more.
Move–anycardwhichisfaceuponthetablecanbeplacedontoanothercardon
thetableprovideditisontheplayingarea.Cardscannotbeplacedontothecards
thathavebeendealt.Therulesthatgovernwhichcardscanbemovedintothe
winpileshavealreadybeenexplained.Withinthedeck,cardmovementis
governedbycolourandnumberofthecardratherthanthesuit.Acardcanonly
beplacedonanothercardthatisnumericallyonehigherthanit,andofthe
opposite(redorblack)colour).Soasevenofclubs(black)canbeplacedonthe
eightofdiamonds(red),butthesevenofhearts(red)couldnot.Norcouldthesix
ofclubs(black),whichistherightcolour,butwrongnumber.
41
Ifcardsareassembledintoastackofcardsinthismanner,theycanthenbe
treatedasasinglecardforthepurposesofmovingthem.Astackthathasthe
sevenofheartsasitshighestcardcanbemovedinitsentiretyontotheeightof
clubs–whichmayitselfbepartofastack.Thesestackscanalsobedisassembled
ifitisdesired,withanycardfromthestackabletoactasthe‘top’cardforthe
purposeofmovingthem.
Cardscannotbemovedasastackuptothewinpiles,sinceobviouslytheybreak
therulesofhowcardsmustbeplacedinthosepositions.
Ifinthecourseofplaythecardthatisupmostandfaceuponapileofcardsis
movedawayfromthatpile,thecardunderneathitisturnedovertobefaceup
andcanthenbeplayedasnormal.
Ifapileofcardsonthetableisexhausted,thatspaceisnowopen.Onlythe
highestcard(aQueen)canbemovedintoiteitheraloneoraspartofastack,and
thequeenscanbeusednormallyfromthatpoint.
Thegameendswheneitherallthecardsareinthewinningposition,thereareno
morelegalmoves,ortheplayerdecidestostopthegame.
Videofootagewastakenofallthegamesplayed.Participantswereinstructedto
playfivegamesofregularsolitaire.Theywereallowedtochoosethepointat
whichtheystopped,eitherwhentheycompletedthegameorfoundthatthey
hadnomoreviablemovestomake.Beforetheystartedtheywerepresented
withstandardrules,andalsoencouragedtoaskanyquestionsthattherulesmay
nothavecoveredtoensureconsistencyofapproach.Thiswasduetothehigh
variabilityofsolitairerulesandtheprevalenceof‘houserules’thatare
commonlyused.Participantswerefilmedwhilstundertakingthetask,and
encouragedto‘thinkaloud’aboutthemovestheyweremakingiftheyfeltthey
hadanythingtonote.
Afterfivegamestheywerethengivenoneoftwoadditionalrulesforthegame:
‐ The‘FreeCell’conditionintroducedafreecellsuchasthosefoundinthe
namesakecardgame,whereanyonecardcouldbeplacedandstoredwith
norestriction(otherthantherequirementthatithadtobepermissibleto
42
movethecardinthefirstplace).Oncethefreecellwasoccupieditwas
full,andnomorecardscouldbestoredthere.Cardscouldberemoved
fromthefreecellatanypointwherealegalmovewaspermissible.
‐ The‘WildSevens’conditionmadeaddedarulespecifyingthat,sevens
wereexemptfromthenormalrulesofcardplacement.Theycouldbe
placedonanycard,andhaveanycardplaceduponthem.However,ifthey
wereusedinamannerthatwouldnormallyhavenotbeenpermissibleno
stackbeyondoneinitialcardcouldbeplacedonthem.
Theserulesrepresenteddifferentmethodsofobtainingfunctionallysimilar
changestothegameenvironment.Inbothconditionsitwasnowpossibleto
movecardsoutsideofthenormalplacementrules.
Finally,oncethegameswerecompleteparticipantswereaskedfortheir
thoughtsontheirstrategieswhendealingwiththenewrulesinaquestionnaire
(seeappendixtwo).
SortingandCoding
Multipleempiricalmeasuresweretakenfromtheexperiment,primarilyfrom
studyingthevideotapesretrospectively.Fromplayback,thetimeofeachmove
madewasrecordedandcodedaccordingtothetypeofactionthatwasoccuring.
Thecodingsusedwereasfollows:
Uncover:Amovewhereacardisuncovered,eitherturningacardfaceuponthe
board,ordealingcardsfromthepack
Normal:Amovewhereafaceupcardismovedontheboardofplay,eitherfrom
thepack,betweencardsontheboard,oruptothefinishingpiles.Anycard
movementthatcouldbeperformedundertheregularrules,andinnoway
involvedthenewrules.
Special:Amovewherethenewrulesarebeingboughtintoplay(movingcardsto
thefreecell,playingacardonawildseven)
Reverting:Amovethattakesacardcurrentlypositionedasaresultofthenew
rules(inthefreecell,onawildseven)andplacesitbackintoregularplay.
43
Pass:Whereanewmovewaspossible,butdeclined.Primarilycodedforwhen
sevensappearedinthedealingpack,butwerenotutilised(sincetheycouldbe
placedonanycard,theyarealwayscapableofbeingused)
NonSpecial:SpecifictotheWildsevenrule,whereasevenisusedinamanner
thatdidnotrequirethenewrule(placedonaneight,forinstance)
Thesecodingswereconsideredsufficienttocovertherangeofpossibleactions
interestingtothisanalysis,withoutbeingsospecificastodefytheidentification
ofanypatterns.
Participants
Therewere6participantsintotal,threeineachcondition.Thereweretwo
womenandonemanineachcondition.Themeanageofparticipantswas24,and
themedian23.
ResultsandAnalysis
Asthenatureofthisstudywasexploratoryratherthanhypothesistesting,the
analysisfocusedonlookingforpatternsandtrendsinthedata.Statistical
analysiswasnotemployedforavarietyofreasons‐primarilyalackofstatistical
powerandanevidentlackofcontrolofextraneousvariablesinparticipant
selection.
OverallData
Twoempiricaloverallmeasuresweretaken.Firstly,thetotaltimetakentoreach
theendofagamewasmeasured.Notethatasitwaspossibleforagivengameto
beimpossibletocompleteorthataplayersimplyfailed,thisdoesnotnecessarily
representa‘complete’state.Inordertomeasurehowcompleteagamewas,a
secondmeasureof‘depth’wastaken.Thiswasacountofhowmanycardswere
faceuponthetableattheendofthegame(includingcardsmovedintofinishing
piles),andthereforehow‘deep’intothegameaparticipanthadmanagedtoget.
Therewasaminimumscoreof0,andamaximumscoreof41.Thiswasbecause
therewerenokingsinuse(makingthemaximumnumberofcards48)andseven
cardswerefaceuponthetableatthestartofthegame.Athirdmeasureof‘time
perdepth’(TPD)wasgeneratedbydividingthetimebythedepthobtainedto
44
giveanindicationofhowquicklyparticipantsweremakingprogressthroughthe
gamewhilstcontrollingfordifferingplaylengths.
FreeCell Wild7
Time
(seconds) Depth
TimePer
Depth
Time
(seconds) Depth
TimePer
Depth
Baseline 228.61 21.93 11.22 387.55 20.53 21.90
RuleChange 257.74 24.27 10.98 467.63 25.13 18.91
RawChange 29.13 2.33 ‐0.24 80.08 4.60 ‐3.00
Percentage
Change 12.74 10.64 ‐2.14 20.66 22.40 ‐13.67
Table3.1:Meanvaluesforoverallmeasures,allfiguresgivento2dp
Thesefiguresprovideanumberofbasicobservationsabouttheexperiment.Post
rulechangebothgroupsappeartobegettingdeeperintothegame.Their
efficiencymayalsohaveimprovedastpdwasdifferentforbothgroups.
Thegroupsmayhavestartedatdifferentlevelsofbaselineskill.TheWildSevens
conditionparticipantstooksignificantlylongeranddidnotgetasfarintothe
gameasintheFreeCellconditionevenbeforetherulechange.Thedifferencein
depthachievedasaresultoftherulechangebetweenthetwogroupsappears
reasonablyconsistentconsideringthesmallsamplesize,thereforethetworules
appearreasonablyequalintermsoftheactualbenefittheygivetoplayersfor
completingthegame.
ConfidenceandPerformance
Priortotheexperimentasnoted,participantscompletedabriefLikertscale
questionnairethatmeasuredparticipantconfidenceintheirplayingabilities.
Thisconcernedbothsolitairespecifically,andcardgamesandstrategygenerally,
includingtheirperceptionoftheirabilitytoadapt.Toexamineiftheremightbea
relationshipbetweenconfidenceandtaskperformance,thesescoreswere
plottedagainstthedifferencebetweendepthscoresbetweenconditions.
45
Figure3.1:Participantconfidence(xaxis)againstrelativechangeindepth
betweentrials(yaxis)
Figure3.1suggestsacorrelationmayexistbetweenthetwofactors.Thismay
simplyrepresentparticipants’accurateestimationoftheirownabilities.
However,itwasalsopossiblethatthiswasindicativeofthemannerinwhich
newstrategiesweretriedandadopted;thatconfidenceindicateshowsuccessful
aparticipantmightbebecausetheyweremorewillingtotrythings.
LearningandProgression
Alsoofinterestinthisstudywashowparticipantsprogressedoverthecourseof
theexperiment;thedevelopmentofstrategiesandskill.Themetricsdetailed
abovecouldalsobetrackedonagame‐by‐gamebasis.Initially,thebaseline
gameswereaveragedasawhole(sinceatthispointtherewasnodifference
betweentheconditions).
46
Figure3.2:Meantimetakentocompletebaselinegamesforallparticipantsin
secondsbytime(yaxis)andgameprogression(xaxis)
Intheabove(figure3.2)thereappearstobealearningeffect,completingthe
taskwithincreasingspeedinlatergames.Howeverasimilarpatterncannotbe
seenforgamedepth(seefigure3.3below)
Figure3.3:Meandepthscoresforallparticipantsinbaselinegamesbygamedepth
(yaxis)andgameprogression(xaxis)
Hereitcanbeseenthatthedepthobtaineddoesnotseemtoincreasewith
increasinggames,asmightbeexpected.Itseemslikelythatthehighvariabilityof
47
potentialsetupsandoutcomesinthetaskisthereasonforthis,butno
conclusionscanbedrawnfromthedata.
Progressionwithintherule‐changetrialswasalsochartedandcompared
betweenthetwoconditions.Fortimingdata,therewasadistinctdifference
betweenthetwo,ascanbeseenbelow(infigure3.4).
Figure3.4:Meantimetakentocompleteamodifiedrulegameinsecondsfor
FreeCellandWildSevenconditions,bytime(yaxis)andgameprogression(xaxis)
Ascanbeseenabove,theFreeCellconditiondemonstratesarelativelylinear
progressiondownwardsthatisconsistentwiththepatternseeninthebaseline
condition.IntheWildSevenscondition,however,thereisanupwardstrendof
timetakenthatcontinuesforbothgames3and4,suggestingthatitisnota
simpleaberration(andtheparticipantdataconfirmsthatthisisnotdowntoa
singleinstanceofapersontakingaverylongtimeandskewingtheaverage).
Thiscouldbeforseveralreasons,themostobviousbeingthatthereissomesort
oflearningprocessgoingonfortheWildSevenparticipantsinthoselatergames
thatisnotoccurringintheFreeCellcondition.Thisiscausingtheparticipantsto
takelongerandpresumablyasaresultgaingreatergamedepth.
48
Figure3.5:MeandepthscoresforFreeCellandWildSevenconditionsbydepth
score(yaxis)andgameprogression(xaxis)
Here(infigure3.5)twothingsareofparticularnote.Firstly,thehypothesis
abouttheincreasedtimebeingasignofincreasedprogressionappearstobe
accurateintheWildSevencondition,providingabasisforlookingforsomesort
ofpatternofnewbehaviourinthattime.Secondly,theFreeCellconditionscores
actuallydeclineovertime–asurprisingresultgiventhattheyalsohadanew
ruletouseandgoesagainstexpectationsofincreasingability.However,itshould
benotedthatallofthegamesotherthanthelastonearestillbeatingthepre‐rule
averagedepthobtained,sothereisstillgeneralimprovementbetween
conditions.Thedeclinecanbeattributedtotwofactorsinalllikelihood:firstly
therandomnatureofthegamesjustbeinglesssolvablebychance,andsecondly
thatparticipantshadbeenplaying10gameofsolitairebytheendofthe
experiment.Itislikelythattherewasadegreeoffatigueatthispoint(severalof
theparticipantsmentionedsuchfeelings)andthedropinperformanceforthe
lastgameinbothconditionsisprobablyindicativeofthisboredomwiththetask.
PatternAnalysis
Havingestablishedfromtheabovedatathattherewasreasontobelievethatthe
twoconditionsmayhavebeenresponsibleforthedifferencesinperformance
metrics,thenextquestionwaswhetherthedifferentrulesmaybegenerating
49
differentstrategicapproaches.Forthisthecodingsnotedearlierwereusedto
examinewhatmovesweremadeatwhattime.
Graphicalrepresentationsofallthegamesplayedweregenerated,anda
selectionofthegraphsandkeyobservationsarepresentedhere.Combination
graphswerecreatedofmultiplegames.Sincethegameslastedadifferent
amountofrawtime,thesewereconvertedintoapercentageandoverlaid.
InitialApproaches
Theplaypatternsofthefirstgameswiththenewrulesforbothconditionswere
comparedandcontrasted.Forthesegraphs,allparticipantgamedatawastaken
andpresentedonanormalisedpercentagescaleoftime.
Figure3.6:Movetype(seekey)playtimebypercentageofgametimeelapsed(x
axis)forthefirstmodifiedrulegameforallparticipantsinWildSevenscondition.
Figure3.7:Movetype(seekey)playtimebypercentageofgametimeelapsed(x
axis)forthefirstmodifiedrulegameforallparticipantsinFreeCellcondition.
Ascanbeseenfromtheabove(figures3.6‐3.7),themoststrikingdifference
betweenthetwoconditionsistherelativetimeatwhichparticipantsusethe
newrule.IntheWildSevenscondition,theruleisonlyusedtowardstheendof
50
thegame,andonlyafterasignificantamountofcyclingthroughthedeck.It
appearsthatparticipantswereonlyusingitafteralltheotheravailableoptions
hadexhaustedthemselves.Thismaybepartiallydowntoavailability;inthisrule
conditiontherewasnoguaranteeaparticipantmighthaveaccesstoasevenat
anygiventime,unliketheFreeCellexample.However,asthechartshowsthere
infactwasavailabilityofsevens,butthattheywerebeingusedasnormalmoves
instead.Itmaythusbeinferredthatparticipantshadtheoptionofmakinga
move,butchosenotto.
Bycontrast,theFreeCellconditionischaracterisedbysignificantlyearlieruseof
themove.Thereissomeevidenceofcycling,asintheWildSevenscondition,
beforeuse,butnottotheextentthatisfoundintheWildSevensinstance.Here
insteaditappearsthatparticipantsbothhavetheoptionofmakingamove,and
arechoosingtomakeuseofit.
Giventhatbothrulesarefunctionallyverysimilar,andthattheyatleastprovide
forthepossibilityofsimilarapplicationthatparticipantschoosenottoutilise,it
canbeinferredthatdifferentstrategiesmaybebeingappliedasaresultofthe
differentpresentationoftherules.However,analternativehypothesiswouldbe
thattherelativecomplexityoftheWildSevensinstancewasdeterring
participantsfromitsuseincomparisontothemorestraightforwardFreeCell.
FinalStrategies
Asabove,thefinalgamesforeachconditionwerechartedaccordingtomove
type.
Figure3.8:Movetype(seekey)playtimebypercentageofgametimeelapsed(x
axis)forthefifthandfinalmodifiedrulegameforallparticipantsintheWild
Sevenscondition.
51
Figure3.9:Movetype(seekey)playtimebypercentageofgametimeelapsed(x
axis)forthefifthandfinalmodifiedrulegameforallparticipantsintheFreeCell
condition.
Inthesecharts(figures3.8and3.9),twothingsinparticularareofnote:
Firstly,thepatternsofusearenowbroadlyidentical.Thisisnotinabsolute
terms,astheWildSevensconditionisusingtherulemore,butintermsof
distributionbothconditionsnowusetheruledistributedthroughoutthe
playtimeofthegames.Withintheindividualgames,agreatdealofcycling
throughcardscanbeseeninallconditionsbeforethespecialruleisappliednow.
Theactualuseoftheruleappearstobelargelyequalbetweenthetwo
conditions.Thissupportstheexpectationthattheconditionswereequivalentin
thefunctionalchangetheywereinducing.
ThereisnowamuchgreateramountofpassingthatcanbeseenintheWild
Sevenscondition.Thisissignificantbecauseitindicatesthatparticipantshave
movedtoavoidingusingthecardsuntilthereisaspecificpurpose.Thisis
suggeststhatthiscondition,liketheFreeCell,hasmovedtowardsusingtherule
asa‘getout’clausewherenoothermoveisavailable.
Thesignificanceofthesefinalgraphsisessentiallythattwodifferentrulesthat
havebeeninitiallyappliedindifferentmannershavemovedtowardsa
consensusapplicationofbeststrategy.
Theseobservationscannotbetakenasempiricalevidenceforthereasonsstated
earlier.However,usingtheobservationsobtainedasastartingpointfor
speculation,twohypothesesforsubsequentempiricalinvestigationwere
generated:
52
1) Strategyformationisinfluencedbycontext,andthemannerinwhich
informationispresented,independentofthesemanticcontentofthat
information
2) Feedbackwilleventuallyovercomethiseffectandtrendtowardsabest
logicalstrategy.
AModelofNovelAdaptation
Inadditiontoprovidingthebasisforfuturehypothesistesting,theabovedata
alsoprovidesabasisforsomecognitivemodeling.Aswiththeabovehypothesis
generationthisisapurelyspeculativestepwhichwillbeaddressedempirically
insubsequentwork.
Fromtheaboveanalysis,andadditionalexaminationofindividualparticipant
gamedata,afour‐stagemodelofnoveladaptationcanbeproposed.Thisisa
high‐levelconstruct,mappingtheoverallprocessratherthanprovidingclose
specificityoftaskimplementation,andforgoodreason.Thenatureofadaptation
toanovelscenariois,bydefinition,goingtoinvolveutilizingspecificprocesses
thatcannotbeanticipatedaheadoftime(elsethescenariowouldbeanticipated
andthusnotnovel).Thus,anabstractlevelaccountofthegeneralizedprocesses
ismoreappropriate.
Figure3.10showsthefourproposedstagesofadaptation.Ascanbeseen,the
modelisacircularprocess,sinceadaptationandtheabilitytoadaptare
propertiesthatseemtobeinnate,andcontinuouslyactivatedforhumans.This
modelassumesthatthisisduetoaperpetualcycle,ratherthanindividual
monitoringsystems,asitseemsunlikelythattheconditionsforasingle
monitoringsystembeingactivatedcouldbesatisfactorilydefinedgiventhe
rangeandscopeoftherealworld.Thefourstagesareasfollows:
Comprehension:Concernedwithwhathaschanged.Thiscoversboththe
processingofinformationtobeabletomakethatdistinction,andalsothe
makingofsaiddistinction.Iftheunderstandingofagivensituationisunderstood
ofthecurrentmentalmodel(aspreviousworkgenerallydoes)thenthisstage
involvesupdatingsaidmodel.Itdoesnotinvolvemanipulationofthe
53
informationwithin,beyondtheprocessofbecomingawareofthechangesthat
haveoccurred.Notethatthisallowsforthepossibilityoferrors.Thesemay
include,butarenotlimitedto,failuretonoteachange,noticingachangewhere
thereisnoneandcorrectrecognitionbutincorrectattributionofeffect.
.
Figure3.10:Ahighlevelcognitivemodeloftheabstractstagesinvolvedin
adaptationtoanovelscenario.
Itencapsulatesboththeformalchangesinrules,butalsotheimplicationsof
theserules.Asanexample,whilsttherulefortheWildSevensconditioncouldbe
verbalized,theexactwayinwhichthisimpactsgameplaymaynotbefully
understood,especiallysincealthoughtheruleitselfissimple,theinteractionsit
producesarenot.Thegradualprocessofcomprehensioncouldbeseeninthe
WildSevenscondition,asparticipantsgraduallyalteredtheirpatternsofuseof
thesevensastheimplicationsofthoserulesbecamemoreapparentthrough
feedback(seelater)
Formulation:Althoughnotdirectlyobservableassuch,thisstageisatheoretical
propositionthatmustexistgivenourunderstandingofcognition.Herethe
alteredsituationisanalysed,considered,andnewapproachesandstrategiesare
constructedonthebasisofthementalmodelproducedinthepreviousstage.
Implementation:Enactingtheplansproducedinformulation.
54
Notethatitispossibletochangeplans,oralterbehaviour–butthisstill
indicatesanothertripthroughtheloop.Implementationintentionallyonly
coverstheliteralenactingoftheplanformed.Thiscouldevenbetosimplytake
noaction.Itisalsopossiblethattherecanbesmalladjustmentstoanongoing
plan–again,thiswouldfallunderthepurviewofbeingproducedinformulation.
Feedback:Theprocessofgatheringinformationaboutthestateoftheworldand
specificallythesystembeingconsidered.
Thisencompassesbothunconsciousinformationbeinggatheredsimplyasa
matterofcourse(tactilefeedback,visualinformation),aswellasanygathering
asaresultofplannedaction.Anyinformationthatanactorbecomesawareof
andthereforeinformsorupdatesthementalmodelisconsideredtobe
‘feedback’,albeitofvaryingsignificance.
Asshouldbeapparentfromthisinitialdescription,thisisanearlyframework
withinwhichtoconsidertheproblemofnoveladaptation.Thereareanumberof
assumptionsunderlyingitthathaveyettobetestedeitherexplicitlyorasa
resultofmorespecificobservationalstudies.Asanexample,itassumesthat
processingproceedsinalinearpathwayaroundthisloop,whereanalternative
hypothesismightspeculatethatfeedbackloopsexistbetweentheindividual
sectionsasappropriate.Thisisconsideredtobeareasonableproposition,since
thestagesarelargelydependentuponeachotherforanyupdating–amental
modelwillnotbeupdatedwithoutnewinformationofsomesort(evenifthat
informationwasalackoffeedback),forinstance.Still,thisisthereasonitexists;
tostartaskingquestionsofthatnature.
Explanations&Predictions
Thismodelcanbeusedtoexplainsomeoftheresultsseeninthisstudy.
IntheFreeCellcondition,participantsunderstoodthechangealmostinstantlyin
termsofhowtousetherule,butnottheimplicationsorthebest‐usepolicies.
Therefore,aperiodofcomprehension,wheretheexactimplicationsoftherule
areexploredin‐situcanbeobserved–participants‘tryingout’therulerightat
thestartoftheirfirstgame.Oncethishasbeencompleted,abetter
55
understandingoftheimplicationsisgainedandwhatcansubsequentlybe
observedisaniterativeprocessofrefinement,viaformulation,towardshowto
applythatunderstanding.
IntheWildSevenscondition,thecomprehensionstagewasmorecomplicated.
Duetothelessstraightforwardnatureoftheruleseveraliterationsoffeedback
informingcomprehensionwererequiredbeforeboththeuseitselfandthebest
usepolicieswereformulatedintoasimilarformtothatseenmorerapidlyinthe
FreeCellinstance.Inbothcasesdevelopingimplementationcanbeobserved,and
feedbackintheformofresultsisconstantlyforthcoming.Thedifferinglevelsof
formulationthatcanbeinferredcanthereforebepresumablyattributedto
differinglevelsofcomprehension.
Discussion
Thisstudysoughttoprovidequalitativeinspirationtobasesubsequent
empiricalworkon.Ithasnotprovidedspecificvalidation,buttheunverified
observationscanprovideabasisforfurtherwork.
Theapproachappearstohavepotential.Bylimitingdowntheproblemspacetoa
solitairecardgame,varyingstrategiesresultingfromdifferentpresentationsof
equivalentrulesmayhaveoccured.Althoughasimplificationofthegeneralized
principlebeingstudied,itseemsthattheapproachisprovidingviableresultsfor
furtheranalysis.Thereforetheprincipleofalimitedbutpotentiallydynamic
environmentwillbecarriedforwardsforsubsequentstudies.
Thestudyalsosuggeststhatcontextappearstobeabletodrivedifferent
strategicapproachestoessentiallythesameproblem.Thishasprovidedthe
basisforaskingnewquestionsatthestrategiclevel,andalsofordefiningthe
boundariesofthenextexperimentstobeabletoaddressthesequestions.
56
CHAPTERFOUR–FramingTaskInstructions:TheCardGameStudy
Introduction
Havingidentifiedtheimpactofframingeffectsonstrategicbehaviourintasksas
anareaforpotentialinvestigation,thischapterwilldetailafirststudyin
addressingthisexperimentally.
Ataskwasdesignedwherestrategicinformationwasframed.Thisreliedupon
theexpandedandreviseddefinitionofframingasdetailedintheliterature
reviewchapter.Thetaskwasintendedtobefamiliarandthusnotutterlynovel,
inordertolimitlearningeffectsasmuchaspossibletobeingattributabletothe
framingeffectsemployed.Acardgamewasusedforthispurpose,basedroughly
ontherulesof‘Uno’althoughmodifiedasdetailedbelow.Theintentwasto
presenttheinstructionsofthetaskintwodifferentways,andsimplybyframing
thesameinformationindifferentwaystocausedifferentstrategiestobe
employed.
Design
Forthepurposesofthistask,ascenariowasrequiredthatwouldberestricted
enoughtoenableaccuratemanipulationoftheframingeffect,whilstalsobeing
openenoughtoenableviablealternativestrategiestoexist.Experiencewiththe
previousexperimentsuggestedthatcardgamesprovidedaneatlylimited
environmentinwhichtoconstructsuchascenario,butthatpattern‐spotting
solitairegamesdidnotprovidesufficientstrategicrangetoaddressthis
question.Consequentlyamultiplayercard‐gamewaschosenasthebasisforthe
design,whereamajorityoftheactionsanddecisionscouldbemeasured,
understoodandexplicitlyquantified,whilestillenablingarangeofoptionsand
tacticalchoicestobeobservable.Theuseofamultiplayergameraisesquestions
ofexperimentalvaliditybutthesewerenotconsideredsufficienttoquestionthe
results(seethecaveatssectionlaterfordetails).
57
Inthissection,therulesofthegameusedwillbeexplainedfirst,andthenthe
mannerinwhichthegameinstructionswereframedwillbedetailed.Afterthese
generalhypothesesforthetaskwillbedescribed,andthenthespecific
experimentalprocedurewillbeexplained.
GameRules
Thegameusedtookthebasicstructureandformofthegames‘Crazy8’s’and
‘Uno’,withsomemodification.Thesewereselectedbothfortherelativelysimple
natureofthegamesconcerned,butalsotheirfamiliarity–theyemploybasic
mechanicscommontomanycardgames.Itwaschosentoframetheinstructions
giventoparticipantsastheindependentvariable,anticipatingthatthiswould
causestrategicdifferencesbetweenconditions.
Therulesofthegamewereasfollows:
Objective:
Thegamewasplayedoveratotaloffourrounds.Theobjectivewastobethe
lowestscoringparticipantattheendoftheserounds.
Eachroundcouldendinoneoftwoways:
1) Oneplayermanagedtogetridofalltheircards
2) Thedeckwasexhausted,andtherewerenomorecardstopickup.
Attheendofaround,pointswereallocatedaccordingtohowmanycardseach
playerhadintheirhand,andwhattypeofcardsthesewere.Allpowercards
countedfortenpoints,whilstallothercardscountedforonepointonly.
Play:
Eachplayerwasdealt6cards.Playproceededinaclockwisedirection.Onecard
wastakenfromthedrawpileandplacedface‐upinthecenterofthetable,which
wastreatedexactlyasthoughithadbeenplayed,includinganyeffects.Each
playerthentookitinturntoplaceacardonthepileinthecenterofthetable.
Thecardhadtobeeitherthesamesuit,orthesamenumberasthecardface‐up
inthecenter.Onlyonecardcouldbeplayedatatime,regardlessoftype.
58
Ifaplayercouldnotplay,orchoosetonotplay,theyhadtodiscardacard
insteadfacedownontothediscardpile.Theythenpickeduptwocards.There
wasanexceptiontothisruleforwhenaplayerhadonlyonecardleft,inwhich
casetheyjustdrewasinglecard.
PowerCards:
Somecardshadadditionaleffectswhenplayed.Thesewerereferredtoas‘power
cards’.
Twos:Thenextplayerhadtopickuptwocards,andthenmisstheirturn.
Jacks:Thenextplayerhadtopickupfourcards,andthenmisstheirturn
Aces:Couldbeplayedontoanything,atanytimeregardlessofsuitornumberof
thecardonthestack.Theplayerthennominatedasuitforthenextplayerto
playto.
Eights:Skippedthenextplayer’sturn
Sevens:Reversedthedirectionofplay.
FrameConstruction
Inorderthatthegameinstructionscouldbeframing,thereneededtobeviable
alternativeapproachestothegametoselectbetween.Scoringcardsattheendof
eachroundprovidedtheseoptions.
Withnoscoringsystemitisunambiguouslybeneficialtobeholdingpowercards
comparedtonormalcards,astheyenabletheplayertohavemoreofaneffecton
theprogressofthegame.Thescoringsystemcausedpossessingthemtobeof
ambiguousbenefit.Theycouldallowmorebehaviour,butwouldpotentially
penaliseaplayerattheendofaround.Havingthedeckrunningoutofcardsas
anend‐conditionaddedanadditionalpressureonthisdimension,astherewasa
finiteamountoftimeinwhichtogetridofcards.
Anotheradditionwasmadetotherulesintheformofdiscarding.Because
playerswereexpectedtowishtogetridoftheirpowercardsbyanymeans
59
necessary,itwaspermittedforplayerstodiscardacardatanytimeratherthan
takingtheirturn.However,asapenaltytheywouldthenhavetodrawtwocards.
Theriskofdrawinganadditionalpowercardshouldnotbeasgreatasthe
potentialbenefitofgettingridofoneyouarealreadyholdingifthegamelooks
likelytoendsoon.Playerswerealsoobligedtodiscardiftheycouldnotplay,and
againhadtodrawtwocards.Thisforcedplayerstomakechoicesaboutwhich
cardstheyvaluedwhichcouldbequantitativelymeasured.
Anovelruleaboutdiscardingwasalsoaddedtoenabledifferentbehaviour.
Becausegettingridofpowercardswasexpectedtobethepreferredstrategyfor
onesideoftheframe,itwaspermittedforplayerstodiscardacardatanytime
ratherthantakingtheirturn,atthepenaltyofpickinguptwocards.Playerswere
alsoobligedtodiscardiftheycouldnotplay,andagainhadtodrawtwocards.
Thismechanismwasexpectedtoshowcardpreferencebetweenthetwotypes
(power/normal).
Framingwasconductedalongtheexpandedlinesdetailedintheprevious
chapter.Itwasbothmulti‐dimensional(comprisingseveralchangesintherule
text)andlesstightlyboundasclassicalstudieswouldrequire.
Theexperimentwasdividedintotwoconditions,highvalueandlowvalue.All
participantswerepresentedwithinstructionsthatexplainedthegameandrules
atthestartoftheexperiment.Theinstructionsemphasiseddifferentstrategies
forwinningviaaseriesofframingmanipulations.Thehighvalueinstructions
emphasisedthatpowercardswereusefulinfacilitatingawin,implyingthatthey
couldbemoreusefulifheldontoratherthanquicklyplayed.Thelowvalue
instructionsemphasisedthepenaltythatplayersfacediftheyheldontopower
cards,andtheideathatitwasmoreimportanttogainanoveralllowscorethan
itwastowinindividualrounds.Semanticdifferenceswerekepttoaminimum.
Someunevenrepetitionwasusedforemphasis,buttherewasnoinformationin
oneconditionthatdidnotexistintheother.
HighValueInstructions:
Todayyouwillbeplayingasimplecardgamewiththreeotherparticipants.
60
Thegamewillbeplayedoveratotaloffourrounds.Theobjectiveistobethelowestscoringparticipantattheendoftheserounds.
Ineachround,theobjectiveistoscoreaslowaspossible.Theroundcanendinoneoftwoways:
1) Oneplayermanagestogetridofalltheircards2) Thedeckisexhausted,andtherearenomorecardstopickup.
Attheendofaround,pointsareallocatedaccordingtohowmanycardseachplayerhasintheirhand,andwhattypeofcardstheyare.
Allcardscountforonepoint,exceptforpowercards,whichcountforten.
Thebestscorepossibleis0,ifaplayermanagestogetridoftheircardsinallfourrounds.
Play:
Eachplayerisdealt6cards.
Playproceedsinaclockwisedirection.Onecardistakenfromthedrawpileandplacedfaceupinthecenterofthetable,whichisplayedexactlyasthoughithadbeenplayed,includinganyeffects.Eachplayertakesitinturntoplaceacardonthepileinthecenterofthetable.Thecardmustbeeitherthesamesuit,orthesamenumberasthecardfaceupinthecenter.
Onlyonecardcanbeplayedatatime,regardlessoftype.
Ifaplayercannotplay,orchoosestonotplaytheyareforcedtodiscardacardinstead,facedownontoadiscardpileinfrontofthem.Theymustthenpickuptwocards.
TheExceptiontothisruleiswhenaplayerhasonlyonecardleftandcannotplay,inwhichcasetheyjustdrawacard.Finalcardsthereforecannotbeexchanged.
Whenaplayerhasonecardleft,theymustsayoutloud‘lastcard’.Failuretodothismeansthattheymustpickupfromthedeck.
PowerCards:
Powercardsarecardsthatcarryaspecialrule.(Theydohoweverincurapenaltyattheendofthegameiftheyareheldattheendofthegame).Importantly,thesecanbeusedtofrustrateotherplayersorhelpaplayersowngame.Theycanskipotherplayersturns,givethemextraturns,orenabletheplayertoplayonanycardtheychoose.
Twos:Thenextplayermustpickuptwocardsbeforetakingtheirgo.
61
Jacks:Thenextplayermustpickupfourcardsbeforetakingtheirgo.
Aces:canbeplayedontoanything,atanytime.Theplayerthennominatesasuitforthenextplayertoplayto.
Eights:Skipthenextplayersturn
Sevens:Reversethedirectionofplay.
LowValueInstructions:
Todayyouwillbeplayingasimplecardgamewiththreeotherparticipants.
Thegamewillbeplayedoveratotaloffourrounds.Theobjectiveistobethelowestscoringparticipantattheendoftheserounds.
Eachroundcanendinoneoftwoways:
1) Thedeckisexhausted,andtherearenomorecardstopickup.2) Oneplayermanagestogetridofalltheircards
Attheendofaround,pointsareallocatedaccordingtohowmanycardseachplayerhasintheirhand,andwhattypeofcardstheyare.
Allpowercardscountfortenpoints.
Allothercardscountonepointonly.
Thescoreattheendofthegamewillbethesumofthetotalsfromtheendofallrounds.
Play:
Eachplayerisdealt6cards.
Playproceedsinaclockwisedirection.Onecardistakenfromthedrawpileandplacedfaceupinthecenterofthetable,whichisplayedexactlyasthoughithadjustbeenplayed,includinganyeffects.Eachplayertakesitinturntoplaceacardonthepileinthecenterofthetable.Thecardmustbeeitherthesamesuit,orthesamenumberasthecardfaceupinthecenter.
Onlyonecardcanbeplayedatatime,regardlessoftype.
Discarding:
Therearetwoconditionsunderwhichaplayermaydiscardcards:
1) Aplayercannotplay2) Aplayercanplay,butchoosestodiscardinstead
62
Inbothcasesaplayerdiscardsacardontoapileinfrontofthem,andthendrawstwocardsfromthedeck.
TheExceptiontothisruleiswhenaplayerhasonlyonecardleftandcannotplay,inwhichcasetheysimplydrawacard.Finalcardsthereforecannotbeexchanged.
Whenaplayerhasonecardleft,theymustsayoutloud‘lastcard’.Failuretodothismeansthattheymustpickupfromthedeck.
PowerCards:
Powercardsarecardswithadditionalrules.Theyalsoincuragreaterpenaltythannormalcardsifaplayerisholdingthemwhentheroundends.Powercardsofalltypescountfortenpointswhenagameends,comparedtoonlyoneforanormalcard.
Twos:Thenextplayermustpickuptwocardsbeforetakingtheirgo.
Jacks:Thenextplayermustpickupfourcardsbeforetakingtheirgo.
Aces:canbeplayedontoanything,atanytime.Theplayerthennominatesasuitforthenextplayertoplayto.
Eights:Skipthenextplayersturn
Sevens:Reversethedirectionofplay.
Thedifferencesbetweenthetwoversionsoftheinstructionsareasfollows:
1) Inthethirdparagraph,thehighvalueconditionincludesanadditional
sentenceabouttryingtoscoreaslowaspossibleineachround.
2) Theorderoftheendconditionsisreversedwhenlisted.LowValue
participantsweretoldthegameendedifthedeckwasexhaustedfirst
(implyingitwasdangeroustoholdontopowercards)whilsthighvalue
participantsweretoldthegameendedwhenoneplayergotridofalltheir
cardsfirst(implyingthewaytowinwastotrytogetridofallyourcards)
3) Thedescriptionofhowcardswerescoredwasbrokenupintotwolines
forthelow‐valueparticipants,andthetenpointvalueofpowercards
highlightedinbold.
63
4) Highvalueparticipantsweretoldthatthelowestpossiblescorewas0,by
winningallfourrounds.Lowvalueparticipantswereremindedthatthe
gamewouldbescoredonthesumtotalfromallfourrounds.
5) Theexplanationfordiscardingisbrokenintolistformforthelow‐value
participants.
6) Powercardpreamble:Inthelowvalueconditionthepenaltythatpower
cardsincurishighlightedinboldtext,acomparisontothecostofnormal
cardsismade,anditisreiteratedthatpowercardscoasttenpoints.Inthe
highvalueconditionitisstatedthatpowercardscanbeusedtofrustrate
anopponent.
Hypotheses
Participantswereexpectedtoshowadifferenceingameperformancebetween
conditions,asmeasuredbytheiroverallscores(cardpoints)attheendoftheir
game.Itwaspredictedthatthehighvalueframingconditionparticipantswould
holdontotheircardsmore,andthusscorehigher.
Itwasalsohypothesizedthatevidencethatparticipantswereutilizingdifferent
strategieswouldbeseeninarangeofadditionalmeasuresrelatedtocarduse
andcontextualchoices.Itwasanticipatedthatparticipantsindifferent
conditionswouldtakedifferentamountsoftimetomakeaplayingdecisionand
willmakechoicesindifferentproportionsasaresultoftheframing.Highvalue
conditionparticipantsareexpecttotakelongertomakeadecisionandtochoose
toplaymorenormalcardsthanpowercardswhenachoiceisavailable.
Finallyitwashypothesizedthatevidencewouldbeseenthatdifferencesin
strategiesreducedovertime,asfeedbackchangedthestrategicapproachbased
onevidence.
ExperimentalProcedure
Inordertomakethegames–andthusthedecisionsoftheparticipants–as
similaranddirectlycomparableaspossible,pre‐sorteddeckswereusedforeach
64
ofthefourroundsofeachgame.Inthisway,thestartinghandsofallplayers
wereidenticalforeachround,thesameplayerstartedeachround,andtheorder
inwhichthecardsweresequencedinthedeckwasalsoidentical–althoughonce
playersstartedmakingchoicesaboutwhichcardstoplaythisinevitablywasno
longercompletelyidenticalbetweenconditions.Itdid,however,ensurethat
cardsgenerallymadetheirwayintothegameatapproximatelythesametime
betweengroupsandconditions.
Arandomcardorderwasgeneratedforeachroundofthegamethrough
shufflingandthatorderwasmaintainedforallconditionsandgroups,withsome
additionalmanipulations.Firstly,decksweresortedsothateachparticipanthad
three(randomised)powercardsandthreenormalcardsintheirhandwhen
startingeachround.Secondly,thedeckswerearrangedsothatthelast6cardsin
eachdeckwerenormal,ratherthanpowercards.Thiswasdonetoensurethat
allplayershadatleastthreepowercardsthatneededplaying,thateveryplayer
startedfromasimilarpositionwhichwasidenticalinscoringvalue,andsothat
asmanypowercardsaspossibleenteredplay(cardsattheendofthedeckwere
lesslikelytobeplayedthanthoseatthestart,sinceaplayercouldwinbygetting
ridofalltheircardsandendingtheround).Playerswereinformedthatthedecks
hadbeenrandomisedinitially,butwerenowidenticalbetweenconditions.
Theexperimentwascounterbalancedsothatthereweretwomalesandtwo
femalesineachgroupinordertoaccountforpotentialgenderdifferences,and
forpotentiallydifferentgroupdynamicsbetweenunisexandmixedgroups.
Participantswereseatedtoplayboy‐girlalternately.Forbothconditionsitwas
alternatedbetweenexperimentalinstanceswhetheraboyorgirlbeganthefirst
game,althoughthecardstheplayersreceivedfortheirinitialhandremained
consistentwiththeirposition.
Beforereadinganyoftheinstructionsorbeingtoldthenatureofthetaskthey
wereundertaking,participantsfilledinabriefquestionnaireconsistingofboth
generalisedwrittenanswerstogaugetheirpriorexperienceandopinionsanda
likertscalemeasureofcompetitiveness,logicalthinkingandgamesplaying
experience(seeappendixthree).Participantswerefilmedwithtwovideo
65
cameraswhilstplayingthegame.Uponpickingupordiscardingtheywere
instructedtoshowtheircardstothecamera,andwerepromptedtodosobythe
experimenteriftheyforgot.Playerswereincentivisedbytheprizeofabarof
chocolateforthewinner,althoughthegeneralattitudeofplayersandtheir
questionnaireresponsesindicatedthattheyweregenerallysufficiently
motivatedbythegameforitsownsake.Scoresattheendofeachroundwere
keptsecret,aswerediscardedcards,sonoparticipantswereawareofhowwell
otherplayersweredoingcomparedtothemuntilthefinalscoreswererevealed,
whichmaintainedmotivation.
Caveats
Thisexperimentwasamulti‐playergame,whichintroducedpotential
confoundingfactors.Thisdecisionwasmadeforanumberofreasonsandthe
resultsarestillbelievedtobevalid.Subsequentexperimentsremovethe
multiplayerfactortoinpartprovidevalidationforthetheoriesproducedinthis
chapter.
Amoreconventionalexperimentalsetupwouldbetosimulatesuchadecision
processinasinglepersonparadigm.Fromthepriorexperimenthowever,there
wasconcernthatalimitedenvironmentwasnotchallengingordynamicenough
torequirestrategisingfromparticipants.Thepresenceofcompetitorsaddressed
thisconcern.Italsohadthebenefitoflendingecologicalvaliditytotheprocess,
whereasrigorouslycontrolledexperimentalparadigmscanbepotentially
artificialandunrepresentative.Itshouldbepossibletosimulateanappropriate
environmentonacomputer,buttherequiredtechnicalskillswerenotavailable.
Inapilotstudyparticipantswereinstructedtoremainsilent,buttheunrealityof
thatenforcedsituationwasobservedtodistractthem.Inthemainexperiment
playersweresimplytoldtoplaythegameastheywouldsocially,andgroups
enforcedacompetitivesilencebychoice.Occasionalcommentswererestricted
tocompetitivebanterthatdidnotappeartoaffectbehaviour,butwereseenas
beneficialfortherealismofthesituation.
Noindicatorsthattheenvironmentwasbiasingwereobserved.Because
participantsheldtheirhandanddiscardchoicessecret,itwasnotpossibleto
66
observeotherstrategiessincethenecessarycontexttounderstandmovesinthat
waywasmissing.Playerswereaskedaboutthisinthepost‐experiment
questionnaireandindicatedthattheyhadnotmimickedotherplayers’choices.
Resultssupporttheideathatframingwasadefiningdifferencebetweengroups
anditisthereforebelievedthattheparadigmwassuitableforthetask,although
asnotedtheseresultswillbesubsequentlyaddressedinamorerigorously
controlledparadigmforvalidation.
CollectionandAnalysis
Resultswerecollectedandanalysedfromthreemainsources.Firstlyoverall
measuresaboutthegamewerecollectedasitwasongoinginordertolookfor
generaltrends.Thisconsistedoftheroundscoresandcards(andcardtypes)
discardedforeachparticipant.
Secondly,Resultsfromthequestionnairewereexamined,andtheLikertscale
questionssortedandtotalled.Qualitativeanswerswerealsoexaminedfor
commentsthatsupported/contradictedthequantitativefindingsoncethe
analysiswascomplete.
Thirdly,videodatawasexaminedinordertoprovideinformationaboutcard
choicesandtimingdata.Forallmovesthatweremadeinallgamesthefollowing
informationwasrecorded:
‐ Thecondition,round,andgroup.
‐ Thecardsthateachplayerwasholdingwhenthemovewasperformed,
thenumberofNormalandPowercardsthatcomprisedthathand,andthe
gamescorethesewereequivalentto.
‐ Theplayermakingthemove,theirpositionandsex.
‐ Thecardfaceuponthestackthatwasbeingplayedupon.
‐ Themovebeingmade(play,discardortake).
‐ Thecardsinvolvedinthatmove,andthetypeofcard(PowerorNormal)
67
‐ Thecontextofthatmove,consistingofwhetherthosedecisionswerethe
onlyoptionoriftherewasanalternative.DiscardingandAceswere
codedinasimilarmanner,buttreatedasspecialcasesasexplainedlater.
‐ Thetimeatwhichthismoveoccurred,andthereforethetimetakento
makethemove.Movesweredeterminedtohavefinishedatthepointat
whichthecardbeingplayedwasplacedonthepile,orthecardsbeing
pickedupmadeitintotheplayer’shand.
‐ Themovepercentage:whichnumbermoveinagameitwas,relativeto
thetotalnumberofmovesinthatgame.
Timingsweremadeaccuratetothenearestsecondusingthein‐builttimerinthe
videorecording.Thiswasconsideredanappropriatelevelofaccuracyduetothe
inherentnoiseininferringcognitivedifferencesfromphysicalaction,andthe
uncertaintyinusingthemtoinfersaidprocesses.
QuestionnaireScales
ThedatafromthequestionnairewassortedaccordingtoscoresforLogical
thinking,gameplayingandcompetitiveness.Thesescoreswereusedtotest
betweengroupstoseeifthefactorsweremorepredicativethanthealternative
conditionsastheanalysisbelowwasundertaken.Noevidenceofthiswasfound.
Additionallydemographicinformationwasusedtosimilarlytestbetweenmale
andfemaleparticipants,butnosignificantdifferenceswereobserved.Thelackof
observableeffectprovideconfidencethatanyeffectsobservedareduetothe
framingeffectsratherthanagreaterdegreeofcompetitivenessorskill.
Participants
Therewereatotalof32participants,evenlysplitbetweenmaleandfemale.
Theyparticipatedinatotalof8groupsof4participantseach,with4groupsin
eachofthetwoconditions.Themeanagewas23.0,andtheMedian20.5.
Results
OverallScore
Theoverallgamescoresforeachparticipantwerenotedandchartedascanbe
seeninfigure5.1.Atwotailedt‐testwasperformedonthisdataandfoundtobe
68
non‐significant.However,thisresultdidnotseemcongruentwiththe
observablevarianceandspread.Thedatawasthereforethencleanedforoutliers
usingtheforthspreadmethod,andthetwohighestresultswereconsequently
removedfromthehighvaluecondition.Anotherindependentsamplest‐testwas
performed(equalvariancewasnotassumed),thistimedemonstratinga
significantdifference,t(25.258)=‐2.120,p=0.044,withLowValueparticipants
scoringsignificantlyhigher(andthereforeperformingworseinthegame)than
HighValueparticipants.
Figure4.1:FinalParticipantScores(yaxis)byCondition(HighValue/LowValue).
Uncleaneddata.
Cleaningthedataappearsjustifiedbytheresultsobtained.Thedifference
observedsuggeststhattheremaybeasystematicdifferenceintheapproaches
takenbetweenthegroups.
PlayingTimes
Timetakentoplaydifferentcardtypes(Power/Normal)indifferentconditions
(HighValue/LowValue)wasanalysed.Thedatawasskewed,asiscommon
withtimingdataanditwasnormalisedbytakingthelogvalue.
69
A2x2ANOVAwithCondition(HighValue,LowValue)andcardtype(Normal,
Power)asbetween‐subjectsfactorsrevealedmaineffectsofcondition,F(1,869)
=13.171,p=.000andcardtypeF(1,869)=5.532,p=.019.Aninteractionwas
notfound.
Figure4.2:Meantimetakenforparticipantstoplayacardinseconds(yaxis)by
CardType(xaxis)andCondition(HighValue/LowValue).Standarderrorshown
aserrorbars.
ThisresultwasthencheckedbyperformingaseriesofMann‐WhitneyUtestson
theindividualelements.Significantdifferenceswerefoundwhencomparing
Powercardsbetweenconditions(U(493)=26865,p=0.026,Z=‐2.228),comparing
normalcardsbetweenconditions(U(379)=15474,p=0.031,Z=‐2.155),and
betweenpowercardsandnormalcardsinthehighvaluecondition
(U(454)=22458.5,p=0.023,Z=‐2.278).Nodifferencewasfoundbetweenpower
andnormalcardsinthelowvalueconditiononitsown(U(418)=18771.5,
p=0.060,Z=‐1.879)althoughitwasclosetosignificance.
ContextualData
Thedecisionofwhichcardtoplaygoesbeyondsimplywhichtypeofcardis
available,butalsoincludeswhetheralternativesareavailable,andwhatthose
are.Havinganalternativeavailableshouldshowadifferenceinstrategicchoice.
Cardusewascodedforeachmove.Movesweredescribedintermsofwhether
70
theothertypeofcard(Normal/Power)wasavailabletoplay.Theywereeither
the‘onlyavailable’,whichmeanthattherewasnocardofthealternativetype
availabletoplay,or‘alternativeavailable’wheretherewasthealternativetype
ofcardavailabletoplay,butwasnotchosen.Thiswasalsodonefordiscarding
movesinthesameway.Aces,however,werecodedseparatelyandtreatedasa
separatetypeofpowercardforreasonsthatwillbediscussedlater;theyarenot
includedinthisanalysis.
Cardchoicewascomparedbetweenconditions.Forallinstanceswherethere
wasadecisiontobemadebetweenplayinganormalcardorapowercardthe
proportionsofthischoicewerecomparedbetweenconditions.Itwasfoundthat
highvalueparticipantschosetoplaynormalcardsinthissituationmore(21%)
thanlowvalueconditionparticipantsdid(12%).AAchisquareresultrevealed
thatthiswasstatisticallysignificant,c2(1,N=303)=4.506,p=0.034.
Figure4.3:Percentageofmoves(xaxis)whereacardtypeischosenovertheother
(Power/Normal)bycondition(HighValue/LowValue)
Thedatawasalsotestedwithinindividualrounds(seefigure).Therewereno
differencesbetweenconditionswithinrounds.Thereweredifferencesbetween
theproportionsinthefirstandlastroundsoverall(c2(1,N=143)=10.93,p=
0.01)),inthehighvaluecondition(c2(1,N=72)=4.055,p=0.044))andalsoin
71
thelowvaluecondition(c2(1,N=71)=5.588,p=0.018)).Allofthismovement
wastowardsplayingpowercardsinpreferencetonormalcards.
Figure4.4:Percentageofmoves(yaxis)whereanormalcardischosenovera
powercardbyround(xaxis)andcondition(HighValue/LowValue)
Additionally,thetimetakentomakeachoicewerecomparedbetween
conditionsusingaseriesofMannWhitneyUtests,butnosignificantdifferences
werefound.Thisisnotinlinewiththeoveralldatafindingsdetailedabove,but
canbelargelyexplainedbyalossofpower.Thisdatasetwassignificantly
smallerthantheoveralldata,anditisnotsurprisingthatthiswouldresultin
insufficientpowertodetectwhatmayhavebeenareal,butrelativelysmall,
effect.AnANOVAandseriesofMann‐WhitneyUtestswerealsoperformedon
themovepercentagedatatoseeifthesedecisionsweremadeatdifferentpoints
inthegames(thistechniqueisexplainedindetaillater)butagainnosignificant
differenceswerefound.
Overall,contextdatasupportstheideathattherearedifferentchoicesoccurring
asaresultoftheframing.
AceData
Aceswerenotcountedasapowercard,eitherasanalternativetoanormalcard,
orcountedwhenplayedasbeingapowercardintheprevioussectiondueto
theiruniqueproperties.Whereallothercardscouldonlybeplayediftheirface
72
orsuitvaluematchedthatofthecardonthetopofthestack,acescouldbe
playedonanycard,regardlessoffaceornumbervalue.Thisenabledagreater
degreeofchoice,andtheexistenceofviablealternativestrategies.Highvalue
participantswouldbepredictedtoholdontoaceswhengivenachancetoplay
them,andlowvalueparticipantswouldbeexpectedtoplayaceswherethey
werenotneeded.
Aceplayingwascodedcontextually.Becauseacescouldalwaysbeplayedoncein
aplayer’shand,therelevantcontextwaswhethertheyweretheonlyoption
availableatthatpointorifadifferentcardwasavailabletobeplayed.This
wouldillustrateifparticipantswereholdingonto,orgettingridofaces.
Thenumbersofeachtypeofthesemoveswerecollectedforbothconditions,and
achisquaretestwasperformedonthedata.Thetestrevealedtherewasa
statisticallysignificantdifferenceinthemoveproportionsbetweenconditions,
c2(1,N=60)=5.610,p=0.018.Inthehighconditionaceswereplayedwhenan
alternativewasavailableonly33%ofthetime.Inthelowconditionthiswas
55%ofthetime.
Figure4.5:Numberoftimes(yaxis)anaceisplayedinagivencondition(Only
Option/AlternativeAvavilible)byCondition(xAxis,High/Low)
Acedatawasalsocomparedbyround.Nodifferencewasfoundforthefirstor
thirdrounds,butwereseeninthesecond(c2(1,N=28)=4.368,p=0.037)and
73
fourthrounds(c2(1,N=24)=5.714,p=0.017),inbothcaseswithHighValue
participantstendingtowaituntilplayinganacewastheonlyoptionmore.There
wasalsoadifferencebetweenroundsoneandfourintheHighValuecondition
(c2(1,N=23)=4.915,p=0.027).
Figure4.6:ProportionofAcesplayed(yaxis)whenanalternativeavailableby
round(xaxis)andCondition(Low,High)
GamePositionData
Anadditionalsourceofdatatobeexaminedforevidenceofsystematic
differencesisthepointinagameatwhichanaceisused.Ifthereisasignificant
differenceinwhenthecardisplayedbetweenconditions,thiscouldbeevidence
ofdifferingstrategies.Rawtimedatawouldbepotentiallyskewedandbiasedby
pausesingameplayordifferinggamelength.Instead,themovecountwasused
togenerateagame‐completepercentagewhichindicatedtherelativepositionof
amoveinthegame.Acemovetypeswerecategorisedinthesamewayasthe
previoussection(AlternativeAvailable/OnlyOption)astable4.1summarises.
A2x2ANOVAwithCondition(HighValue,LowValue)andmovetype
(Alternative,OnlyOption)asbetween‐subjectsfactorsrevealedamaineffectof
MoveType,F(1,104)=3.970,p=0.049andnoothereffects.Thistest,however,
failedtomeettheexpectationsofhomogeneityandthereforeaseriesofMann‐
WhitneyUtestswereundertakentocomparebetweenthegroupsdirectly.A
74
significantdifferencewasfoundbetweenmovetypesintheHighValuecondition
U(52)=184,P=.027,Z=‐2.214butnotintheLowValuecondition,suggestinga
differenceinaceusebetweenconditions.
High Value Low Value
Alternative
Available 70.0 54.60
Only Option 54.50 50.50
Table4.1:Meannumberofgamemovescompleted(percentage)whenanAceis
playedbycondition(HighValue,LowValue)andMoveType(AlternativeAvailable,
OnlyOption)
Figure4.7:Frequencydistributionofproportionoftotalpopulation(yaxis)oftime
takentoplayacard(xaxis)whenanalternativeisavailableintheHighValue
condition.
Theindividualmovetimedataforthissetwasthencompared.Sincethedata
wasnon‐parametricaseriesofMann‐WhitneyUtestswereusedtotestbetween
theconditions,andasignificantdifferencewasfoundbetweenthetimetakento
playanAcewhenthereisanalternativeavailablebetweenthehighandlow
conditions(U(48)=172p=0.048,Z=‐1.976).Nosignificantdifferencewasfound
formovetypewithinthehighvaluecondition,althoughitwasclose(U(52)=210,
75
p=0.087,Z=‐1.711).TheAlternativeAvailabledatawasthenplottedas
frequencydistributions.
Figure4.8:Frequencydistributionofproportionoftotalpopulation(yaxis)oftime
takentoplayacard(xaxis)whenanalternativeisavailableintheLowValue
condition.
Discarding
Afinalsourceofdataisthecardsdiscardedinthecourseofthegame.Discarding
wasbothaconditionofbeingunabletoplay,andalsowasavailableforplayers
tochoosetodoinlieuofamove.Intheoryitprovidedanotherwaytobeableto
getridofpowercards.
Thisdatawascodedinthesamewayastheothermoveswere,withitbeing
notedifcardswerediscardedwilfullyornormally,whattypeofcardtheywere
andalsoifthecardtypediscardedwastheonlyavailable,ornot.Aseriesoftests
wereperformed,butnosignificantresultswerefound,eitherasameasureofthe
proportionsofchoicesmadeineachconditionforwhichachisquarewas
performed,norforthetimingdatabetweenconditionsandcardtypesforwhich
aseriesofMann‐WhitneyUtestswereundertaken.Therewerenoidentifiable
differencesbetweentheconditions.
76
Figure4.9:Percentageofdiscardedcardsthatwerepowercards(yaxis),byround
(xaxis)andcondition(HighValue,LowValue)
GraphingthedataasinFigure4.9suggeststheremaybeapotentiallearning
effect.Nosucheffectcouldbestatisticallydetected,althoughthedatasetwastoo
smalltoconsiderthisconclusive.
Discussion
Thehypothesesdescribedatthestartofthissectionwillbeevaluated,andthen
additionalconclusionsthatemergedasaresultoftheevidencegathered
detailed.Implicationsforthemodelspecifically,andthenthemoregeneral
significancewillbeevaluated,beforenewhypothesesaresuggestedforfurther
study.
HypothesisTesting
Thefirsthypothesisstatedthatadifferencewasexpectedtobeobserved
betweenconditionsasmeasuredattheoverallgamescorelevel.Thiswas
supportedwiththecaveatthatthedatarequiredcleaningforoutlierstodetect.
Secondly,itwashypothesisedthateffectswouldbeseenatadditionalmeasures.
Thiswassupported,aseffectswereseenbetweenconditionsforcardchoice
(power/normal)timetakentoplay(power/normal)andwithintheaceplaying
data.Thedifferenceswerealsofoundinthedirectionthatwasanticipated,with
77
highvalueconditionparticipantschoosingtoholdontopowercardslonger(both
overallandforacesspecifically)andtakinglongertomakeadecisionoverall.
Finallyitwashypothesisedthatdifferencesbetweenconditionswouldreduce
overtime.Thiswasnotsupportedasdifferencesremainedincardplaying
choicesbetweenthefirstandlastroundforbothnormal/powercardchoices,
andalsowhenaceswereplayed.However,inbothcasestherewasevidencethat
thetrendwasinthesamedirectionbetweenconditions.Thissuggeststhatthe
underlyingideabehindthishypothesis(thatfeedbackwouldinformanoptimal
strategycommontobothconditions)mayhavesomevalidity.
AdditionalConclusions
Firstly,wecanconcludethatthedifferenceobservedinoverallscoresisrealand
significant.Therangeofsystematicdifferencesdemonstratedinthemove‐level
analysesprovidessufficientevidencetobelieveparticipantswereacting
differentlyinthedifferentconditions,andthedatacleaningremovedoutliers
thatwereskewingthedataandobscuringarealeffect.Participantswere
successfullyframedbydifferentinstructions,withhighvalueparticipants,on
averagescoringbetter.Wecanseethisinthedifferenceoftheoveralltiming
data,thedifferenceinthechoicetousenormalorpowercardsandintheuseof
aces.
Secondlythereappearstobeaninteractionbetweenframingandfeedback.
Althoughthecardchoiceresultsdidnotreachasharednormandtheconditions
remaineddifferent,theydidmoveinthesamedirectionwhichsuggeststhatboth
setsofparticipantswereadjustingtheiractionsinthesamedirection.Itwould
appearthataframecanbealteredbyfeedback,althoughthereisinsufficient
evidencetoconcludewhetheritwouldeventuallyrenderaframeirrelevant.
Additionallyitshouldberememberedthatthisresultisnotnecessarily
transferabletoothertasks.Thefeedbackinthisscenariowasunambiguous,
accurateandcomplete.Intherealworlditisentirelypossibleforfeedbacktobe
misleading,ambiguous,inaccurateorincomplete–anditisuncleartowhat
degreethismayhaveanimpactonframing.
78
Havingdrawnconclusionsfromsomeofthespecificresults,moregeneral
conclusionslookingatthebodyofresultsasawholewillbeaddressed.
SelectiveApplicationofFramingEffects
Thereisevidencethatwithinagiventaskthesameframingcanhavedifferent
resultsondifferentaspectsofthattask.Thisimpliesthatonsomeleveltheframe
isconsideredandappliedaccordingtothecontextadecisionismadein.Itcanbe
easytothinkofframingeffectsassettingalargelyunconscioustendencywhich
isthenconsciouslyapplied–aviewpointlargelysupportedbythefactthat
participantsdonotgenerallyacknowledgebeingframed,evenwhenitis
illustratedtothem(Gigerenzer,2000).Inthisexperimentevidencesuggeststhat
framingisnotnecessarilyautomaticorgenericbutappropriateandcontextual.
Inthisexperiment,measurementsofframingeffectsweretakenforthree
measures;a)betweennormalandpowercards,b)theuseofacesandc)
discarding.Differenttypesofeffectwereseenateachdespitethefactthateach
involvedelementsofthesamebasicdecision:howandwhentoplaypower
cards.Forthechoicebetweennormalandpowercardstheeffectwastoalterthe
ratioofthedecisionmade,butitseemsnottheunderlyingstrategyorrationale
behinditastherewasnoevidenceofdifferencesintimetakentoplayorwhen
thedecisionwasmade.ForAces,notonlywasthechoicemadedifferently
betweenconditions,buttimingandpositioningdatashowedthatthiswas
varyingaccordingtothecontextitwasmadein.Finally,fordiscardingnoeffect
couldbeobservedatall.Whencontrastingbetweenthesethreedifferentresults
itiscleartoseethatthesameframeishavingadifferenteffect(ornon‐effect)in
threedifferentaspectsofthesametask.
Apossibleexplanationforwhythisisoccurringcanbeworkedfromprior
research,whichobservedthatambiguityappearstohavearoleindetermining
whenandhowaframingeffectisapplied.Theremustbealevelofuncertaintyas
totheoptimalcourseofactionorcorrectanswer(ascanbeseenintheAsianflu
examplecitedearlierwherethestatisticaldifferencebetweenthesolutionsis
nil).Inthesesituations,participantsrelyonframinginformationtodiscriminate
79
betweenotherwiselargelyindistinguishablechoices.Thesuggesteddifference
hereisthatwhenambiguityvaries,framingbehavesdifferently.
Whenchoosingbetweenpowerandnormalcardsthelevelofambiguityand
uncertaintyislow;powercardsaregenerallybestplayedwhentheopportunity
presentsitself,astheobjectiveisultimatelytoberidofallcards,andtheycarry
morepenaltythanmostwhilstprovidinglittlebenefitfromholdingontothem.
Thereareplausiblebenefitsincertainsituations,andgenerallyitisaquestionof
whetheraplayerwantstotaketheriskofholdingontoone,buttheweightof
evidencesupportsgettingridofthem.Theframingeffectislimitedbecausethe
amountofambiguityislimited,andtheeffectequatestothatofapplyingthe
samediscriminatoryheuristicindifferentparameters,namelyholdingonto
powercardsonetimeinten,oronetimeintwenty.
Incontrast,whenconsideringtheuseofAcesthereisagreatdealofambiguity.
Therearecompeting,plausibleconsiderations:ononehandtheyrepresentthe
samedangerasanypowercard,inthattheywillsignificantlyincreasea
participant’sscoreattheendofaroundandthereforeshouldbegottenridof
quickly.Ontheother,theyarealsoabletobeplayedonanythingandtherefore
areespeciallyusefulwhenaplayerhasveryfewcardsandismorelikelytofacea
situationwherenoneoftheircardscanbeplayed.Asaresultofgreater
ambiguitytheframingeffectismorepronouncedinitseffect,notonlyintermsof
whichchoiceismadeinadecision,butalsothestrategyonwhichthatdecisionis
basedisdifferent.Hereinthelowconditionparticipantsarefarmoreeagerto
getridoftheirAces;theyplayfarmoreoftheminsituationswhereanothercard
isavailable,andtakealotlesstimetothinkaboutitwhentheydo.Bycontrast
thoseinthehighvalueconditionnotonlytakemoretimebeforemakingsucha
move,buttheyalsomakethosemoveslateroninthegame,atatimewherea
strategyisactuallyviableoruseful.Sointhiscaseparticipantsareapplyinga
differentheuristicaltogether,dependingoncondition.
Whatthisexplanationdoesnotaccountfor,however,isthediscardingdata.Here
ambiguityisatleastsufficientthatthereisuncertaintyaboutthebesttacticto
proceedwith–morepowercardsarediscardedasaproportionofthetotalin
80
laterrounds,andthepossibilityistheretotreatcardsdifferently.The
expectationwasthatthiswouldbeframedliketheotherelementsofthetask.So
thequestionbecomes:whynot?
Inthistask,participantswerepresentedwithasituationwherethebasic
mechanics(powercards,aces)werelargelyfamiliartothemasaresultofhaving
playedsimilargamesbefore,somethingsupportedbythequestionnaire
responses.Discarding,however,wasnotpartofthisgenerallyfamiliarparadigm,
asevidencedatleastinpartbythefactitwasnoteasilyunderstoodatfirst–it
wascommonfortheexperimentertoneedtocorrectorexplaintheprocedure
earlyon,andseveralerrorsconcerningthiselementofplaywereretrospectively
notedwhenthevideowasexamined.
Thistiesinwiththeexplanationsuggestedforthedifferencesinframing
betweentheothertwolevels.Aswasnotedprior,there,framingwasoccurring
tosomedegreeatthestrategiclevel.Fordiscardinghowever,participantswere
stillstartingtounderstandtheprocessatall.Noframingeffectoccurredbecause
participantsdidnotunderstanddiscardingwellenoughforittobeinfluenced:
theysimplydidnotmaketheconnection.
However,discardingisnotsocomplexanideathatwewouldexpectpeople
unabletograsptheconceptasanormalprocess–indeed,inpriorresearch
peoplehavebeenframingdoingunfamiliartasks.Thisiswherethesecondpoint
comesin.Becausethereweredecisionsandstrategiestobeimplementedwithin
anunderstoodframeworkandthesecomprisedthemajorityofthetask,thistook
precedenceoveranalysinganewcomponent.Thisfitswiththepre‐existing
researchtheoryoflimitedsearch:thereisafiniteamountofprocessingthatcan
occurbeforeadecisionismade.Hereparticipantslackedthefreecognitive
resourcesnecessarytocomprehenddiscardingsufficientlythatitcouldbe
framed.
Theseobservationstakentogethersuggestthatframingoccurstosomedegreeat
astrategiclevelofthinking,thatplansaremadeandexecutedatleastpartially
basedonthebiasesthataframecanimpart.Thedifferencesinframingeffect
betweenthelevelssupportsthis–asimpleblanketruleof‘trytothrowpower
81
cardsaway’doesnotexplainthatvariation.Ifaccurate,thisprovidesareason
thatitisnecessarytomovebeyondsingle‐decisionparadigmsforstudying
framingeffects,astheyareinsufficienttocapturethepotentialcomplexitiesof
thephenomenatheyinduce.
Finallytheseobservationscreateanumberofimplicationsforexistingtheory.
Firstlytheysupporttheideathatambiguityiskeyinaframingeffectworking,
andextendsthatfindingtoamoretask‐orientatedenvironment.Wecanseethis
bothfromwheretheframewasapplied,butalsoinhowtherewasregression
towardsasharednormbasedonquantitativefeedback.Secondlybywidening
thescopewithinwhichframingeffectsandtheirimpactisconsidered,questions
areraisedabouthowandatwhatpointincognitionaframecanhaveaneffect.
LearningImplications
Thisexperimentprovidessomeinsightintohowlearningmayprogresswhena
framingeffectispresent.Itshouldbenotedatthispointthatthereareseveral
reasonstointerpretthelearningdatawithsomecaution.Playersonlyengagedin
fourroundsofthegame,alimitednumberwhenassessinghowtheparticipants
improved.Thenatureoftheroundsalsoprovidedalimitedamountof
opportunitiesforfeedbackintheformofaroundscore,andthereforealimited
opportunitytoassesshowwellaparticularstrategywasprogressing.
Additionallytherewasvariationbetweenrounds–theconditionsplayedthe
samegamesinthesameorder,butthosegamesweredifferenttoeachother,and
somemayhavelentthemselvestocertaindecisionsorscoresmorethanothers.
Thepointoftheseobservationsisnottosaythatwecannotdrawtrendsfrom
theinformationcollected,butrathertonotethatitislikelyinappropriateto
drawtoospecificconclusionsfromthevariations.Asanexample,boththeAce
usedataandthediscardingdatahaveroundswithslightlyoddpatternsfrom
oneorbothoftheconditions.
However,andwiththesethingsborneinmind,therearestilltwoconclusions
thatcanbedrawnfromtheevidence.Firstly,allthreemeasures(discarding,Ace
useandpower/normalcardchoice)appearedtoshowsomesortofeffect,and
82
secondlyallconditionsalsoappearedtoshowatrendtowardsasharedposition,
ratherthandiverging.
Thesearesignificantbecausetheysuggesttwothingsaboutlearninginafarmed
task.Firstlyitsuggeststhatfeedbackandevidencecanovercomeaframing
effect.Itsupportsthepositionthatitisambiguitythatallowsaframingeffectto
haveaneffect–asfeedbackremovesambiguity,sothestrategyalters.Indeed,
bothconditionssuggestchangeastheroundsgoonforallmeasures,indicating
thatevenwhenthe‘correct’hypothesisandstrategyisprimed,participantsare
stillrefiningandperfectingtheirbehaviouraswouldbeexpected.
Secondly,thisimprovementandthesetendenciesarenotableforthefactthat
theyappeartomovetowardsasharednorm.Inallmeasures,bothconditions
movetowardsthesamepointasanoptimalsolution(whetheritisanoptimal
solutionisopentodebate,buttheconvergenceisstillrelevant).Thetaskwas
shortenoughthattheconvergenceisnotconclusivelyshowntobecompleteby
thetimethegameended,sothereisdefinitelyacasetobemadeforframing
effectsimpactingthetimetakentoreachanoptimalsolution.However,the
overallconclusionisstillpertinent:framingeffectshavenotshownthesolution
todiverge,andthesuggestionthereforeisthatintaskswherethereisanoptimal
statetobereached,framingwillnotaffecttheeventualconclusionreached
providedthereissufficientfeedback.
ModelImplications
Theobservationsinthisexperimentbroadlysupportthemodelasproposed.The
differenceinframeimplementationbetweenthedifferentactivitiesmeasured
canbeexplainedbythestages.Beforethegamebeganandwhenparticipants
werereadingtherules,theycomprehendedthebasicsofthegamefrom
recognizingsimilaritiestoprevioussuchgames.Themainchangesweretothe
scoringsystem,whichaffectedthevalueofthecards,andtheframingprovided
thebasisforthedifferingstrategiesthatweredevelopedasaresult.Sothepre‐
gameformulationproducedframedstrategies,andthereweredifferent
approachestodifferentcardswithinthesestrategies.Feedbackfromboththe
roundscoresandcardsbeingplayedtheninformedthesestrategiesby
83
enhancingcomprehension–primarilytheimportanceofgettingridofpower
cardsifitwasnotinitiallyfullyunderstood.
Thediscardingthatfailedtobeframedcanbeexplainedbythefactthatitwas
notsufficientlycomprehended.Theerrorsthatcamewiththediscardingsystem
supportedthisviewofincompletecomprehension.Itwasonlywiththefeedback
fromplayingthegamethatgreatercomprehensionoftheroleofdiscardingin
thegamecouldbeappreciated,andasaresultonlythenthatformulationcould
produceappropriatestrategiesfordealingwithit.
Thisre‐assessmentofthemodelisshowingsomepotentialpredictivevalidity,
howeveritremainsmoreusefulretrospectively.Questionsthatarisefromthis
experimentareinhowtopredictwhatmaybecomprehendedandunderstoodto
enableformulation–withoutthisknowledgethemodelwillremainmoreuseful
asapost‐hocassessorofeffortstoadaptratherthanpredictingsuccessor
anticipatingpotentialproblems.
Significance
Havingdiscussedtheimplicationsofthesefindingsontheory,andthepotential
futureresearchquestionsitraises,itisworthalsoconsideringwhatsignificance
theseresultsholdasawholefortheoryandeverydaylifeingeneral.
Possiblythemostimportantconclusiontotakeawayfromthisexperimentisthe
apparentlimitsofframingeffects.Itiseasytoreadtheliteratureandassume
thatoureverydecisionandmovemaybedictatedbythemannerinwhichwe
experienceinformation–anassumptionthatfliessomewhatinthefaceof
everydayexperience.Itisnotthatweareunaffectedbyframing,butthatthese
effectstendtomatteratthestartofaprocess.Experienceandfeedbackallowus
toovercomethesecognitivebiased.Framingeffectsarenotforever.
Ofcoursewhilstwesaythat,itisimportanttoconsiderthatsuchasituation
actuallyrepresentsagreatdealofthethingsweencounterandtaskswe
perform.Theexperimentdetaileddescribesasituationthatwasfundamentally
familiar,butwithsomenewelementstoit–adifferentsetofexpectationsonthe
samebasicmechanisms.Thesamedescriptioncouldeasilybeappliedtodining
84
atanewrestaurant,drivinganewvehicleorevenmeetinganewperson.Andas
wecanseefromthedatacollected,framingeffectscanhaveadistinctimpacton
howweapproachatask.Theinitialstrategy,thefirstapproach–theseare
commonlynon‐trivialexperiences,especiallyinsituationswheretheremaynot
betheopportunitytogaintruefeedbackbeforeneedingtoperform–combat
situationsforasoldier,forinstance.Allthetrainingintheworldultimately
cannotaccountforalivefireexperience.Additionally,whilstthefeedbackinthis
caseappearedtocorrectforthesub‐optimalformulationofstrategies,this
occurredmainlyin‐betweenroundswheninformationaboutscoresbecame
apparenttoparticipants.Combatisagoodexampleofasituationwhereuseful
feedbackisoftenofthissort–availableoncetheimmediateexperienceisover
andactionscanbeconsideredatlength,butnotnecessarilyavailableatthetime
andinthechaosofthemoment.
Inthesesituations,thisresearchraisesquestionsaboutthenatureof
instructionsthatshouldbegiven.Specificity,itwouldseem,isimportantnotjust
ingeneraltermsforwhatthesituationis,butinhowthatinformationneedsto
beapplied–thelackofaframingeffectforthediscardingchoiceillustrateshow
havingthebigpicture,eveninthecorrectmanner,manynotbeenough.Smaller
elementsthatshouldbetreatedinthiswaymayfallbythewayside,andto
participants’detriment(discardingwasapotentiallyveryeffectivemeansof
gettingridofpowercards).Thesamesortofconsiderationistruewhendealing
withexceptions.Intheexperiment,Aceswereauniqueformofpowercard‐but
theywerestillapowercardandlow‐valueconditionparticipantsappliedthe
framingsub‐optimallyasaresult.Aspriorresearchhashighlighted,eliminating
ambiguitywouldappeartobekeyforgooddecisionmaking.Itmaybethat
providingpeoplewithgeneralrules–butthenmakingapointofhighlightingthe
importanceofconsideringindividualinstances–isofgreatereffectiveness.
Ofcourse,thosepointsmaketheassumptionthatambiguitycanbeeliminatedto
somedegree,whichisnotnecessarilythecase.Theremaynotbethetimetodo
so,orindeedtherequisiteinformationavailable.Additionally,itmaybethat
framingeffectscanmakegooduseofsuchambiguity–ifacourseofactionis
preferred,itprovidesawaytomakeittheinitialapproachpeopleadoptwithout
85
seemingbiasedintheinstructing.Ifinstructionsarebiasedorperceivedtobeso,
thereisalwaysthechancethatpeoplewillgoagainstthemonprinciple.Indeed,
therearemanysituationswherepeoplemaynotnecessarilytrustthesource
providingtheinformation–intelligentlyframingtheinstructionscouldprovidea
mechanismforinfluencingthepathtakenwithoutseemingtoleadthem.
Thisexperimentalsosuggeststhatitmaybemoreimportanttoframeforthe
criticalthingswherepeoplemaygowrongifambiguityishigh.Thesmaller
thingswhereambiguityislow–inthiscasebasicpowercardsandnormalcards
–won’tbeasaffectedasaresultofthatlowerambiguity.Thekeymayliein
identifyingpotentialpitfalls,andframingforthem.Indeed,thelackofaneffectin
discardingdemonstratesthattherearelimitedthingsthatwillbetakeninwhen
preppingforanewtask.Makingsurethattheessentialpointsaren’tpartofthat,
andhighlightingtheunfamiliar,aresimplestepsthatcouldpotentiallyenhance
performanceincriticalareas.
NewHypotheses
Havingaddressedthehypothesesandexaminedthetheoreticalandpractical
issuesassociatedwiththeseresults,thenextnaturalquestionishowtofurther
thisresearch.
Thisexperimentoffersanumberofobservationsthatcouldprovidethebasisfor
futureinvestigation,anditwillnotbepossibletopursuethemall.However,itis
possibletoidentifysomeofthekeyquestionsraisedtotakefurther.
Firstly,therearegeneralquestionsabouttherepeatabilityoftheseresultsina
differentparadigmandwithdifferentframingmechanismsetc.Whilstthereis
certainlyreasontobeconfidentthatthedatapresentedhereisvalid,thesesorts
ofresultsandspeculationneverthelessdemandconfirmationthatthereisn’t
simplyanordereffectorsituationalbiasatplaythatiscreatingtheobserved
properties.Forthisexperimentandlineofresearchthismeansprimarilythat
anotherexperimentshouldseektoexaminesimilarquestionsutilisinga
differentframingmechanism,andtodosoinadifferentsettingtothatofcard
games.
86
Secondly,thereareanumberofquestionsthatarisefromissuessurrounding
feedback.Thisexperimenthadtwonotablepropertiesconcerningfeedback:a
significantsourceofitwasonlyreceivedattheendofeachround(intheformof
aparticipant’sscoreforthatround)andthatitwasqualitativeandunambiguous
–higherwasworse,lowerwasbetter.Whilstneitherofthesewere
inappropriate,itisalsoeasytoconsideralternativemethodsinwhichfeedback
mightbeobtained;itmightbequalitativeorquantitative,itcouldbereviewed
moredirectlyinlinewithactionstaken,anditcouldbeambiguous.Itiscommon
inreallifetoseeasinglesetofstatisticsinterpreteddifferentlybydifferent
partiesaccordingtotheirprioritiesandbeliefs,soiffeedbackwereambiguous
wouldthesameoptimizationbeseeninatask?Orwoulditsimplybeusedto
confirmthepre‐existingbiasesengenderedbytheframe?
Thirdly,howframingmightpersistalsocreatesquestionsofwhetherthewayin
whichaframeispresentedcanaffectthis.Inthistaskparticipantswereinitially
framed,butnotsubsequentlywhilstundertakingthetask.Itisnothardto
imagineasituationunderwhichframingmightbereinforcedhowever–and
wouldtherepetitionofaframebesufficienttocontinueitsinfluenceontask
decisions?Orwouldfeedbackstillendupasamoreimportantdeterminant,and
repetitionsimplyweakentheeffectoftheframeasitslackofactualpredicative
validitywasshownthroughfeedback?
87
CHAPTERFIVE–FramePositioningandConfidenceinaDescriptiveTask:TheFestivalStudy
Introduction
Previousworkhasestablishedframingeffectsexhibitingdistinctpropertiesin
anongoingtaskparadigm.Thegoalofthischapteristofurtherinvestigatethis
byreportingastudyofframingeffectsinaforcedchoiceexperiment.The
experimentutilizesaprogressingthematicallylinkedseriesofdecisions
conveyingastoryaboutarockfestival.
Thisstudywasdesignedtotestthehypothesesgeneratedbytheprevious
experiment,butalsomovebeyondtheparadigmthusfaremployed(card
games).Itwasalsodesignedtomorecloselyreflectclassicframingapproaches
byreapplyingtheAsianflustatisticalweightingandemployingaforcedchoice
paradigm.Confidencewasalsointroducedasameasureinordertoreflect
graduationindecisionmakingnotreflectedinapureeither/ordichotomy.
Priorresearchhasmadeuseofconfidencetostudydecisionmaking.Participants
havebeenshowntoapproachtasksdifferentlyasaresultofvaryingconfidence
frompriorcontemplation(Koehler,1991),confidencewasseentoaffect
influenceingroupdecisionmaking(Zarnoth&Sniezek,1996)andconfidence
canalsoactasapredictoroftasksuccess(Feather,1968).Hereitwasmeasured
forthesereasons,butalsoasafactorpotentiallyinfluencedbyframing.Itwas
anticipatedthatframedconfidencemaybehigherthanunframed,asthe
(perceived)additionalinformationwouldhelptojustifychoicesmade.
Design
Overview
Ataskconsistingofaseriesof11binaryquestionsthatrelatedtoanongoing
taskwasdesigned.Anarrativewasconstructedaroundtheparticipantmanaging
amusicfestival,duringwhichanoutbreakofavirulent,unidentifieddisease
88
occursandstepsmustbetakentocontrolit.Thebinarydistinctionusedwasone
ofdecidingbetweenusinginternalorexternalsolutions,withtheexactnatureof
thesevaryingaccordingtoquestion.Additionally,aconfidencemeasurewas
takenforeachquestionona7‐pointscale.Onceadecisionwasmade,feedback
wasgivendetailingtheeffectofthatactiononvariousmetrics(money,public
opinionandcasesofdisease).
Twoindependentvariablesweremanipulated;framepositionandfeedback
type.Thetaskwasframedeithertowardsinternalorexternalsolutions,inoneof
threeways.TherewaseitherNoFrame(NF),aframepresentedinthe
instructionshenceforthreferredtoasanOverallFrame(OF)oraframe
presentedinthequestionsthemselvesreferredtoasaQuestionFrame(QF).
FeedbackwaspresentedaseitherNumericdata(N)suchasmoneylost,oras
qualitativestatements(Q)ofthesamething.
Allfeedbackwaspredeterminedandtherewasalinearprogressionthroughthe
task,butparticipantswereinformedthattherewasanelementofchancetotheir
performance,thattheremightnotbea‘good’solutiontoadilemmabutrathera
‘leastbad’oneandthattheiractionscouldaffectlaterevents.Thiswasto
increaseecologicalvalidity.Additionally,allparticipantsunderwentcertain
pretestingmeasuresandcompletedpost‐assessmentquestionnairesassessing
authoritarianism,extraversionandrisktaking(seelaterinthischapterfor
details).
Questions
11questionsweredevisedthatplottedtheoutbreakofapotentiallydangerous
diseaseatafestival.Thefirstfivequestionswererelativelyinnocuous,dealing
withaspectsofsimplefestivalmanagement–lateacts,facilitiesbreakingdown–
whilstthelatter6dealtdirectlywiththegrowingoutbreakproblemandthe
variousissuesthataroseasaresult,eitherdirectlyorindirectly(Seeappendices
thirteen,fourteenandfifteenforexamples,ortheaccompanyingCDforthefull
text).
Eachdecisionwasdesignedsothatthechoiceswerevalidinbothdirections–
internalandexternal.Thesolutionsproposedwereselectiveandasrealisticas
89
thestrictdivisionforcinganeither/orchoicebynatureallowed.Theywere
designedtobeseenasplausibleandviableandthiswastestedthroughpiloting.
Allofthequestions’choiceswerederivedfromreal‐worldoccurrencesand
procedures.Mostimportantlythequestionsweredesignedtobeperceivedto
matterandhaveconsequences.Tofurtherthisend,theywereintentionally
descriptiveinnature;contextualinformationthatdidnotnecessarilyrelateto
thechoiceitselfwaspresentedtogiveasenseofplaceandimmersion.
FramingEffects
TheframeusedwasadirecttranslationoftheAsianfluexampleusedby
KahnemanandTversky(Tversky&Kahneman,1981).Thiswaschosenbothfor
thereasonsthattheycited(statisticalequivalence,reversibilityetc)andalso
preciselybecauseithasbeenrepeatedusedanddemonstratedasrobust.Inthis
case,externalsolutionswereusedconsistentlyasthe30%chanceofsuccess,
andinternalasthe1/3guaranteedtobesaved,whicheverdirectiontheframe
wasappliedin.Forcounterbalancingpurposestherewerebothinternally
favouringandexternallyfavouringframesforquestionframeandoverallframe
conditions.Bothdirectionswereframedidentically,orascloseaswaspossible
withgrammaticallimitationsandconsiderationsofphrasing.
Theoverallframewaspresentedaspartofthetaskinstructionsandwas
describedasageneralproperty–thatexternalsolutionswouldsucceed33%of
thetimeorthatinternalsolutionswouldsave1/3ofwhateverwasatrisk.For
theQuestionFrametheframewaspresentedaspartofthedecisiondescription
attheendandputacrossinthetermsofthequestion(“33%ofthebeerwillbe
saved”)
Feedback
ThreeMetricsoffeedbackwerepresentedforquestions:‘Money’,‘Reputation’
and‘Cases’.Thesewerechosenasimportanttoafestival(moneyandreputation)
orrelatingtothetaskproblem(cases).MoneyandReputationwerepresentfor
theentireexercise,whereasCasesonlystartedoccurringoncetheinitialcases
hadbeenreportedinquestion4.Apatternoffeedbackwaspre‐determinedtobe
asimplepatternofone‐then‐the‐other,inthatfirstaninternalsolutionwouldbe
90
‘correct’andgivepositiveresults,andthenanexternalonewould.Thefirst
questionwasanexceptiontothis,rewardingeitherchoiceequally,ensuringthat
bothframesstartedoffbeingrewardedinthesamewayregardlessofchoice.
Themetricswerebalancedsothatanapproximatelyequalamountof‘money’,
‘reputation’and‘cases’wasawardedortakenawayforinternalorexternal
choices–thatisthatifaparticipantweretousesolelyoneortheotheroption
theywouldreceivethesameamountofpositiveandnegativefeedback,andalso
thesameintensity.Foreachquestionasetamountof‘money’and‘reputation’
wasavailableandthegainorlossofthesedependedonthechoicemade.For
‘reputation’thiswasexactlybalancedwithgainsandlossesvaryingbetween1
and3%foreachquestionequallyforinternalandexternalovertheset.For
‘money’,thegainsandlosseswerebalanced,butnotperfectlyasabove–rather
slightlyrandomnumberswereused‐£4004and£4037ondifferentquestions
forinstanceapproximatelybalancingeachotherbutgivingtheimpressionof
beingrandomlygeneratedtosustaintheillusionofchanceinthetask.
Additionally,therewereapairofquestions(7&8)wherebothchoiceresultedin
numericallosses,withonelosinglessthantheother.Caseswastreatedslightly
differently,inthattheycouldnot‘lose’casesoncetheyhaddeveloped,butrather
thereweresimplydifferentlevelsof‘newcases’presented,andthesewere
similarlybalancedbetweenconditions(seeappendixseventeen).
Theabovedescribesthenumericalcondition.Forqualitativefeedback,the
numericalamountsweredescribedasSlight,Average,SignificantorVery
SignificantGainsorLosses(seeappendixsixteen).
Programming
TheexperimentwasdesignedandruninMicrosoftVisualStudio2010.
Participantsusedanetbookandmousetonavigatethroughthetask.Aninitial
participantnumberboxwasfilledinbytheexperimenter,butfromthispoint
untiltheendofthetaskparticipantshadcompletecontrolofthetask.
Eachquestionwaspresentedasaseparatepage,withthedescriptivetextabove
apairofbuttonsfortheinternalorexternalsolution,withastandard,brief
descriptionofthatchoiceonthebuttons(‘Waitforcar’or‘Startwalking’for
91
example)andaseven‐pointradiobuttonLikertscaleforexpressingconfidence
belowthat.Participantshadtoclicka‘continue’buttontoproceed,andcouldnot
dosountiltheyhadmadeaselectionofchoiceandconfidence(seeappendix
eighteen).Afeedbackpagewasthenpresented,consistingoftwobuttons–‘get
feedback’and‘continue’.Clicking‘getfeedback’wouldcausetheappropriatetext
toappear,andsimilartothequestionpageitwasimpossibletoproceedwithout
clickingtogetfeedbackfirst(seeappendixnineteen).Everyclickperformedin
thetaskpromptedalineofcodetobegeneratedinauniqueexcelspreadsheet
foreachparticipant.ThisdetailedtheQuestionnumber,Pagetype,Eventtype,
Timeandcurrentchoices(forinternal/externalandconfidence).
Pretesting
Eachparticipantunderwentpretesting,fillinginthreestandardised
questionnairesthatmeasuredIntroversion/Extraversion(Francis,Lewis,&
Ziebertz,2006)(seeappendixnine),Authoritarianism(Kelman&Barclay,1963)
(seeappendixeight)andRiskTaking/impulsiveness(Patton,Stanford,&
Barratt,1995)(seeappendixseven).Theseweretakentotestiftherewere
personalitytraitsthatwouldbeconfoundingfactorsfortheexperiments.They
werecontrolledforpost‐hocasitwasimpracticaltodosobeforehand.
Authoritarianismwasmeasuredbecausethetaskpresentedthepossibilityfor
simplydeferringto(outside)authorityoneachchoice.Risktakingwas
controlledforbecauseonesetofchoicescouldpotentiallybeseenasthe‘safer’
option(presumablytheoutsideauthorities).Introversion/Extraversionwasnot
predictedtobeaconfoundingvariable,butratherwastakentoserveasa
baselinenon‐predictivemeasure.
Aftertheexperiment,participantsfilledinabriefquestionnairethattheir
confidenceaboutthetask,whichtypeofsolutiontheyfelttheyhadusedmore
andwhichfeedbackmetrictheymadethemostuseof.Additionallytherewerea
seriesofopen‐endedquestionsabouthowtheyfelttheirperformancehadbeen,
whattheymighthavechangedandtheirgeneralperspectiveonthetask.
92
Grouping&Participants
Theexperimentwasdividedintoequalgroupsbasedonthetwofeedback
conditions(NumericandQualitative)andthreeframingcondition(Neutral,
Over‐FrameandQuestionFrame).Inbothoftheframedconditions,groupswere
counterbalancedtoaccountforframedirection(InternalorExternal)which
addedanadditionaltwogroupsforeachoftheseconditions,makingatotalof
tengroups.
60participantsweretested,withsixineachofthetenconditions.Eachgroup
wasgender‐balancedwiththreemenandthreewomenineach,andallthe
groupswereagebalancedasparticipantswerecollected.Agesrangedbetween
63and19,withamedianof23andameanof26.1.Therewerenosignificant
differencesbetweengroupsonthebasisofage.
Hypotheses
Theexperimenthadthreehypotheses.
Firstlyitwaspredictedthatparticipantsdecisionswouldbeaffectedbyframing.
Boththeoverallframeandthequestionframewouldinfluencethedecisions
made,althoughtheoverallframewouldbelesseffectiveforlaterquestions.
Itwasalsopredictedthatframingwouldaffectconfidence,withframed
decisionsbeingmoreconfidentthanunframed.
Finallyitwaspredictedthatfeedbacktypewouldnotaffectthedecisionsmade
orconfidence.
Results
Theinformationgatheredwasanalysedinturn,primarilyconsistingofthe
pretestingmeasures,decisionsmade,confidenceratings,timingdataandafew
additionalmiscellaneoustests.
PretestingMeasures
Thescoresforeachofthepretestingmeasures(Authoritarianism,
Introversion/Extraversion,Risk)weretotalledandsortedbysubjectand
experimentalresults.Thepopulationofsubjectswasdividedintothetophalf
93
andbottomhalfofresultsforeachmeasureandthesegroupswerethen
comparedaccordingtoexperimentalresults.
Usingconfidenceastheindependentvariable,aseriesofMannWhitneyUtests
wereperformedandnodifferencewasfoundforanyofthefactors.Additionally,
correlationswereperformedfortheaverageconfidenceandpretestingscores.
Norelationshipswerefound.
Whentestingfordifferencesindecisionmaking,aseriesofchi‐squaretestswas
performed.Nodifferencewasfoundfortheintroversion/extraversionmeasures,
orauthoritarianism.Therewas,however,adifferenceinthechoicesmadefor
risktaking,c2(1,N=660)=5.605,p=0.018.
Moreriskadverseparticipantschoseaninternalsolution46%ofthetime,whilst
moreriskseekingparticipantschoseinternal37%ofthetime.Thisindicatesthat
participantssawtheinternalsolutionsasbeingmoreriskythanexternalones,a
reasonablefindinggiventhatthesituationisdepictedasbeingaseriousonethat
wouldnaturallyattracttheattentionofoutsideauthorities.
Asaresultofthisfinding,aKruskal‐Wallistestwasperformedonthe
distributionofrisktakingscoresbetweenthedifferentgroupsintheexperiment.
Itwasfoundthattherewasnosignificantdifferencebetweengroups.This
indicatesthatwhilstrisk‐takingdoeshaveanimpactonthedecisionmade,this
particulartraitwasdistributedevenlythroughtheexperimentpopulation.Asa
result,itcanbeassumedthatalthoughitisapotentiallybiasingfactor,risk‐
takingdoesnotpredictoraccountforanyofthedifferencesobserved.
Decisions
Thedecisions(internal/externalsolution)madebyallparticipantsweretaken
andsortedaccordingtoexperimentalgroupanddecisionmade.Aseriesofchi‐
squaretestswerethenperformedtocomparebetweengroupsforsignificant
differences.TestswereperformedusingtheNoFramedataasabaselineto
compareagainstforallgroups,butwerealsoperformedcomparingbetween
typeoffeedback(NumericandQualitative)anddirectlybetweenframingtypes
(OverallFrame,QuestionFrame).TheproportionsintheNeutralconditionwere
94
50externalto82internal–approximately62%internal.Thisshowsthatdespite
theattempttopre‐balancethequestionstherewasanunderlyingtendency
towardsinternalsolutionsbeforeanyframingoccurred.
NodifferencewasfoundbetweenconditionsforthechoicesmadeintheOver‐
Framecondition,eitherbetweenthetwodifferentframesorcomparingeitherto
theNeutralcondition(Figure5.1).
Figure5.1:NumberofChoicesmade(yaxis)bycondition(xaxis)andchoicetype
(Internal/External)forOverallFrames
SignificantdifferenceswereobservedintheQuestion‐Frameconditionbetween
thetwoframetypes(c2(1,N=264)=45.019,p=.000),andbetweentheinternal
frameandtheNeutralcondition(c2(1,N=264)=7.174,p=.007),andthe
externalframeandtheNeutralcondition(c2(1,N=264)=17.519,p=.000).
Additionally,thepercentagechangeofchoiceswascalculatedforbothframed
(thepercentageshiftindecisionsgiventheNeutralconditionasaperformance
baselineandthereforethedifferentamountofdecisionsthatitwaspossibleto
changegiventheinitiallopsidedness)andwasfoundtobeverysimilar–40%
fortheinternalframeand41%fortheexternalframe(Figure5.2).
95
Figure5.2:NumberofChoicesmade(yaxis)bycondition(xaxis)andchoicetype
(Internal/External)forQuestionFrames.
Thedatawasalsoanalysedbysplittingtheresultsaccordingtofeedbacktype,
butnosignificantdifferenceswerefoundeitheroverallorwithinindividual
conditions.
Theseresultswerethenchartedbyquestion(Figure5.3).
Figure5.3:Percentageof‘external’choices(yaxis)byquestionnumber(xaxis)
andconditiontype(External/Internal/Neutral)forQuestionFramedata.
96
TheOver‐Framedatawasalsographedinthismanner,asshownbelow(figure
5.4).
Figure5.4:Percentageof‘external’choices(yaxis)byquestionnumber(xaxis)
andconditiontype(External/Internal/Neutral)forOverallFramedata.
Comparingbetweenthetwographs,itappearsthattheframingeffectis
consistentintheQuestionFramecondition.IntheOverallFramecondition,there
appearstobenoisierdataforthefirstsixquestions,andrelativeconsensusfor
thelastfive.
Statisticalnote
Itisnecessaryatthispointtotakeamomenttoexplaintheapproachtakenin
regardstosomedataanalysisinboththischapterandthenext.Anoccasional
issuewithdataobtainedforthisthesisisthatthedatafromvarioussourcesdid
notfittheassumptionsofnormalityandhomogeneitythatarerequiredfor
parametricstatisticsevenaftertransformingthedata.
Theexperimentaldesignwasadirectfitfora3x2ANOVAandtheredoesnot
existaconsensusoverthebestwaytoanalysenon‐parametricdatainthisform.
AKruskal‐Wallistestiscommonlyreferredtoas‘thenon‐parametricANOVA’,
butitisonlydesignedforanalysisalongonedimension.TheMann‐WhitneyU
testisawidelyusednon‐parametriccomparisonofmeans,butbringswithitthe
riskofmakingtype1errorswhenrepeatedlyapplied
97
Whereatransformationdidnotresolvetheissueofassumptionsbeingviolated,
anANOVAisperformed,buttheseviolationsnoted.AseriesofMann‐WhitneyU
testsarethenreportedalongthedimensionsoftheANOVAtocomparethedata
setmeansindividually.
InthiswayitisbelievedthattheuseofANOVAswillhopefullyaccountforfalse
positiveresultsfromthemassofMann‐WhitneyUtestsdespitetheviolationof
homogeneity,andtheuseofMann‐WhitneyUteststhattheANOVAresultwas
notinaccuratedespitefailingtomeetassumptionsofhomogeneity.Byengaging
overseveraldimensionsandshowingconsistentresultsitisbelievedthatthe
resultsaredemonstratedtobesufficientlyrobustfordiscussion.
ConfidenceData
Havingexaminedthedirectchoicesmade,theconfidencescoresforeach
participantwerethentakenandsortedbyframetypeandfeedbacktype.The
ANOVAfailedthetestofhomogeneitywithrawscores,andnodata
transformationscouldsuccessfullynormalisethedata.
A3x2ANOVAwithFrame(Neutral,Over‐Frame,QuestionFrame)andFeedback
(Qualitative,Numeric)asbetweensubjectsfactorsrevealedmaineffectsof
Frame,F(2,654)=6.397,p=.002,ηp2=0.19,andFeedbackFrame,F(1,654)=
11.897,p=.006,ηp2=0.12.Thesewerequalifiedwithaninteractionbetween
FrameandFeedback,F(2,654)=26.446,p=.000,ηp2=0.5.
AseriesofMann‐WhitneyUtestswereperformedtocomparebetween
individualgroups.Comparisonswerelimitedtocomparingbetweengroups
whichhadatleastonefactorincommon(frameorfeedback),ascomparing
wherebothweredifferentwouldintroducetoomuchvariabilitytotheprocess.
SignificantdifferenceswerefoundbetweentheNeutralframeandtheOver‐
Frame(U=14108.5,p=.002,Z=‐3.165)andQuestionFrameandOver‐Frame(U
=30870.5,p=.019,Z=‐2.348),howevernodifferencewasfoundbetween
NeutralandQuestionFramesoverall.
98
Figure5.5:MeanConfidence(yaxis)presentedbyFrame(xaxis)andFeedback
Type(Numeric,Qualitative)forallparticipants.Errorbarsshowstandarderror.
Nosignificantdifferencewasfoundcomparingbetweenfeedbacktypesoverall,
butsignificantdifferenceswerefoundbetweenfeedbacktypeswithinall
conditions‐Neutral(U=1284,p=.000,Z=‐4.165),Over‐Frame(U=7526.5,p=
.049,Z=‐1.966)andQuestionFrame(U=6497.5,p=.000,Z=‐3.726).
Withinthenumericfeedbackconditiondifferenceswerefoundbetweenthe
NeutralandOver‐Framegroups(U=2884.5,p=.000,Z=‐3.97),andbetween
theNeutralQuestionFramegroups(U=2434.5,p=.000,Z=‐5.224).However,
nodifferencewasfoundbetweentheOver‐FrameandQuestionFramegroups.
Withinthequalitativefeedbackconditiondifferenceswerefoundbetweenthe
QuestionFrameandNeutralgroups(U=3303,p=.004,Z=‐2.86),andQuestion
FrameandOver‐Framegroups(U=6020,p=.000,Z=‐4.500).Nosignificant
differencewasfoundbetweentheNeutralandOver‐Framegroups.
Theconfidencedatawasalsoanalysedonaquestion‐by‐questionbasis,butdid
notrevealanynotableresults.Thereweredifferencesbetweenindividual
questions,aswouldbeexpected,buttherewerenointeractionwithother
factors.Confidencewasalsotestedaccordingtohowconfidencevariedasa
resultofthechoicemade–inwhetherparticipantsweremoreconfidencewhen
goingwithaframethantheywereagainstit.Therewasnodifferencefound
99
betweenthegroupsforeitherframedgrouporoverall,norwasthereany
interactionwithfeedbacktype.
TimingData
Aspartoftheexperimentalsetup,timingdatawascapturedforeverymoment
whenparticipantsclickedabuttonorotherwiseinteractedwiththeprogram.
Thisdatacanbeusedgiveanideaofthethinkingtimesinvolvedinaproblem,
butisnotstraightforwardtouseforseveralreasons,andnecessitatedtheuseof
severalnormalisingtechniquesthatwillbeexplainedfirst.
Background
Theexperimentalsetuphadparticipantsexposedtoanentirepagewhen
consideringaproblem.Becausetherewasnodistinctionbetweentheperiodin
whichparticipantswerereadingtheproblemdescriptionandtheperiodin
whichtheywerethinkingaboutwhichchoicemightbemoreappropriate,there
wasnodirect‘thinkingtime’metric.Timingdatawas,however,availableforthe
entirepage,thatisthetimetakenfromthepointatwhichtheparticipant
enteredthepagetothepointatwhichtheyleft.Thistimewastransformedfor
thepurposeofstatisticalcomparison.
Readingtime,speedandtextlengthwereallcontrolledfor.Readingspeedwas
assessedbyusingtheinstructionssheetasabaseline.Thetimetakentoreadthe
instructionswastaken,andthendividedbythenumberofwordsreadforeach
condition(astherewereadditionalwordsintheOver‐Framecondition
instructions,andtheQuestionFrameconditionquestions).Thisproducedatime
perword(TPW)metricthatcouldthenbedirectlycomparedbetween
conditions.
Anindependentsamplest‐testindicatedthatscoresweresignificantlyhigherfor
theOver‐Framecondition(M=275.91,SD=107.74)thantheunframed
instructions(M=225.24,SD=63.34),t(34)=2.29,p=.045,d=0.58.Levene’s
Testindicatedunequalvariances(F=4.33,p=.042)sodegreesoffreedomwere
adjustedfrom58to34.Nodifferenceswerediscoveredbetweenfeedbacktype,
aswasexpected.
100
Figure5.6:TimePerWord(yaxis)byframetype(xaxis)andfeedbacktype
(Numeric,Qualitative).ErrorBarsshowstandarderror.
TheresultssuggestthatparticipantsintheOver‐Frameconditionarespending
moretimethinkingabouttheinstructionstheyhavereceived,andnotsimplyas
anartifactofhavingreadmorewords.Secondly,theNeutralandQuestionFrame
conditionsarestatisticallythesameforreadingspeed,suggestingthatthereisa
reasonableexpectationthatthegroupsareaptforcomparisonandtheOver‐
Frameresultisnotsimplytheresultofnaturalvarianceinreadingspeeds.
Thesescoreswerethereforeusedasthebaselineforareadingand
comprehensionspeed.Thattherewasalsocomprehensionoccurringwasnot
consideredaproblem,sincethelevelofnewinformationwasthesameforall
participants,excepttheOver‐Framecondition.Inordertoprovideabaselinefor
thatcondition,theaveragedifferencebetweentheOver‐Frameandallother
scoreswastaken,andtheOver‐Frameresultsreducedbythatamount.
TPWscoreswerethengeneratedfromthequestionpagesforallparticipantsby
dividingthetimetakenbythenumberofwordsinthatgivenquestion.Thisgave
anideaofthinkingtimeforeachquestion,butwerestillpotentiallydistortedby
differentreadingspeeds.Hence,therawTPWscoresfromtheinstructionswas
usedasabaselinetoprojectanexpectedspeed,andthedifferencebetweenthis
score(QuestionTPW)andthebaselinescorewascalculated.Thisdifferencein
101
scorewasthenconvertedtoaproportionofthebaselineTPW,anditwasthis
valuethatwastested.
Analysis
Figure:5.7ProportionChangeinTPWbetweenInstructionsandQuestions(yaxis)
bycondition(xaxis)andFeedbackType(Numeric,Qualitative).ErrorBarsshow
standarderror.
Becausethedatawasconvertedintoapercentage,itwasnotappropriateto
performanANOVAonit.AseriesofMann‐WhitneyUtestsshowedsomeeffect
however.TheNeutralandOver‐Frameconditionsweredifferent(U=15102,Z=
‐2.16,p=.031),asweretheOver‐FrameandQuestionFrameConditions(U=
31327,Z=‐2.01,p=.045).NodifferencewasfoundbetweentheNumericand
QuestionFrameconditions.Therewasnodifferencebetweenfeedbackforthe
NeutralandOver‐Frameconditions,buttherewasonefortheQuestionFramed
condition(U=7340,Z=‐2.212,p=.027).Acorrelationwasfoundbetween
percentageTPWchangeandconfidencebySpearman’srho(r=‐.086,N=660,p
=.027)suggestingthatwhenparticipantsweremoreconfidenttheytookless
timetomaketheirdecision.
Spearman’scorrelationalsofoundarelationshipbetweenpercentageTPW
changeandquestion(r=‐0.174,N=660,p=.000)indicatingthatasparticipants
completedthetasktheytooklesstimeconsideringtheiranswers.
102
Figure5.8:MeanPercentageTimePerWordChangefromBaseline(yaxis)by
Questionnumber(xaxis)
Finally,acorrelationbetweenconfidenceandquestionnumberwasnotfound.
FrameDirectionEffects
Asnotedpreviously,theframedirectionhadaneffect,inthatparticipantswere
lesslikelytochooseexternalsolutionsonageneralbasis,asseeninthebaseline
condition,andthatrisk‐adverseparticipantschosemoreexternalsolutionsthan
internal.
Again,thedatawasinpercentageformandthusinappropriatetobeassessedby
anANOVA,butitwasgraphed,andaseriesofMann‐WhitneyUtestsperformed
onit.
Significantdifferenceswerefoundbetweenconditionsforanexternalframe(U=
6983,p=.005,Z=‐2.787)andbetweenframetypeswithintheOver‐Frame
condition(U=6150,p=.000,Z=‐4.13).Adifferencewasalsofoundbetween
ExternalandInternalframesoverall,(U=27853,p=.000,Z=‐3.991).
103
Figure5.9:MeanPercentageTimePerWordchangebetweenBaselineand
Questionmeasurements(yaxis)bycondition(xaxis)andFrameDirection
(Internal,External).ErrorBarsshowstandarderror.
FeedbackTime
Timingdatawasalsogatheredforthefeedbackdata.Feedbackwasobtainedby
clickingabuttontoseeit,andthenanothertoproceedonceread.Thishadthe
effectofisolatingtheperiodinwhichparticipantswereconsideringthe
feedback,removingtheambiguitythatexistedindistinguishingbetweenreading
timeandthinkingtime.Thelimitedamountofinformationpresentedas
feedbacksimilarlyeliminatedtheneedforreadingtimestobeaccountedforas
therewerenolongstringsoftexttobereadandunderstood.
Therawtimedatawasthereforetaken,andconvertedintoitslogarithmicform
toensurehomogeneityofvariance.A3x2ANOVAwithFrame(Neutral,Over‐
Frame,QuestionFrame)andFeedback(Qualitative,Numeric)asbetween
subjectsfactorsanalysing(logarithmic)timerevealedaninteractionbetween
FrameandFeedback,F(2,654)=4.177,p=.016,ηp2=.013.
104
Figure5.10:MeanLogarithmofFeedbacktime(yaxis)byFrameType(XAxis)and
FeedbackType(Numeric,Qualitative).ErrorBarsshowstandarderror.
AseriesofMann‐WhitneyUtestswereperformedcomparingbetweenthe
individualdatapoints.Asignificantdifferencewasfoundbetweenthenumeric
andqualitativescoresattheQuestionFramelevel(U=6759,p=.002,Z=‐
3.148).Neitheroftheotherlevelsshowedasignificantdifference.
Discussion
Astheprecedingsectionhashopefullymadeclearthisstudyhasprovideda
wealthofdatatobeanalysedandunderstood.Thissectionwillseektotie
togetherthevariousanalysesandmakeclearwhatoverallconclusionscanbe
made.
Firsttheinitialhypothesesthatweresuggestedwillbeevaluated.Followingthis,
specificconclusionswillbedrawnaboutthedataandwhatitseemstotellus
abouthowintegrationoccurs,andalsohowconfidenceinteractswithframing
effectsovertime.Ongoingimplicationsfortheproposedmodelwillthenbe
discussed,beforenewquestionsthatthisstudyhasraisedwillbeaddressed.
Hypotheses
Thethreehypothesescannowbeassessed.
105
Firstlyitwaspredictedthatparticipantsdecisionswouldbeaffectedbyframing.
Thiswassupportedinthecaseofthequestionframe,butnosignificant
differenceswereobservedintheoverallframecondition.Theinconsistencyof
choicesinthefirstsixquestionsinthatconditionpossiblysuggestsaneffectofa
differenttypehowever.
Itwasalsopredictedthatframingwouldaffectconfidence,withframed
decisionsbeingmoreconfidentthanunframed,andfeedbacktypewouldnot
affectthedecisionsmadeorconfidence.Itwasfoundthatframingdidaffect
confidence,butasaninteractionwithfeedback.Theresultsprovedtobemore
complicatedthanthehypothesespredicted,andthiswillnowbediscussedin
moredetail.
Complexity
Beyondthespecifichypothesestherewasabasicexpectationthatframingwould
haveaneffectinthisenvironmentasitdidinthecardgameparadigm.Thiswas
stronglysupported.Effectswereseenbetweenframingconditionsandbetween
feedbacktypesfordecisionsmade,confidencescoresandtimingmeasures.
Howeverasnotedabovetheresultsdifferedfromwhatwaspredicted.Framing
wasonlypresentatthequestionlevelandfeedbackhadanimpactonconfidence
andtimingmeasures.Theresultscontainanumberofinteractions,suggesting
thatoneofthemainconclusionsthatcanbedrawnisthatofcomplexity,andthat
theresultsinthisareagobeyondsimplecausalrelationship.
Theresultsfromtheconfidencemetricaresignificantinthisobservation.Firstly
becauseitisapotentially‘hidden’attribute–thatitwasobservedherebecause
itwaslookedfor,butmayalsohavebeenafactorinpreviousworkandgone
undetected.Thisprovidesevidencethatframinghaspropertiesthatneedtobe
understoodintermsofanongoingscenario,andalsoinbroadertermsthan
simplytheeither/orchoicedichotomytheyhavegenerallybeenstudiedin.
Secondlyitshouldnotbeconcludedthatthisisthereforethetotalityofhow
framingcanaffectthedecisionmakingprocess,butratherevidencethatframing
couldpotentiallyimpactonmultiplefactorsbeyondjustthedecisionmade,of
whichconfidenceappearstobeone.Resultsfromtimingdatasupportthisidea,
106
thatdifferentcognitiveprocessesareoccurringasaresultofframingand
interactionswithfeedback.Therelativesimplicityoftheeffectinsingle‐choice
examplesdoesnotextendtoamoredynamicsystem.Andthisexperimentwas
relativelycontrolledandconstrained–itmustbeanticipatedthatmore
complicatedrelationshipscouldexistinscenarioswithstillgreaterecological
validity.
FramingPositionEffects
ContrastingtheneutralandQuestionFrameconditions,aneffectisreadily
apparent‐theNeutralconditionprovidedabaselinereactionthattheQuestion
Frame‘stretched’inthedirectionofthatframe.Whatwasnotexpectedwasthat
itmaintainedthatpullovertheentireexperiment–therewasnoevidenceof
framefatigue,orparticipantsbecoming‘wise’totheconstantframetheywere
beingexposedto.Fromthepreviousexperiment’sresultsitwasanticipatedthat
overtimeparticipantswouldrelymoreonfeedbackthantheframe(astheframe
wouldbecometransparentlybiasedoncerepeatedenough)andreverttowards
theNeutralbaseline,buttherewasnoevidenceofthis.
Additionallyitwasexpectedthattheframewouldbemostpotentfortheinitial
problems,butthiswasagainnotthecase–ifanythingittookacoupleof
questionsfortheframetostartsignificantlyaffectingdecisions.Giventhatthis
framehasbeeneffectiveinasingle‐decisioncontextpreviously,itseemsmost
likelythatthisinitialfailureoftheframeisduetotheongoingscenario
paradigm.Whatappearstobeoccurringisthattheframeaspresentedtakes
timetobeintegratedintotheoveralldecisionmakingapparatus(providingyet
moreevidencethatthiswas,infact,seenasawholetaskratherthanaseriesof
individualdecisions).Thisexplanationwouldalsoaccountforthenon‐failureof
theframelateron,whenrepetitionwouldhavebeenexpectedtolessenits
effects.
TheOver‐Frameconditionfailedtoframeparticipantsatall,butdidappearto
haveaneffect,astheinitialsixquestionchoiceswereunpredictablebetween
conditions.ThiswasdespitetheQuestionFrameparticipantsmirroringthe
baselinepattern,andover‐frameparticipantsshowingthatsamemirroringlater
107
on(alsoimplyingthattheeffectfadedwithtime,consistentwithitbeingaresult
oftheframe).TheeffectoftheOver‐Framemaythereforehavebeentodisrupt
thedecisionmakingprocessforthefirstfewquestionsofthetask.Thequestions
thisraisesiswhywouldthishappen,andwhynoframingoccurred.
Itishypothesisedthatthiswasduetonewinformationaffectingthedecision
makingprocess,butnotbeingclearenoughtobesystematicallybiasingor
framing.Ifwebelievefromthequestionconditionthatittooktimeforthebiasto
befullyintegratedintothedecision‐making,thenitwouldfollowtheconverse
couldbetrue,andinthiscasetheframewasneverfullyintegratedand
understood.Inadditiontoonlyoccurringonceasopposedtorepeatedframing
(asinthepreviousexperiment)theframewasbynecessitypresentedinabstract
termsof‘externalorinternal’solutions.Theinformationwasnotcontextualised
anyfurtherorlinkedtoanythingspecificinthequestions.Sowhilstparticipants
wereawarethattherehadbeenadditionalinformationimparted,theymayhave
beenunsureexactlyhowtoapplyit.Theymayalsohavebeenconfusedabout
whichsolutionwas‘internal’andwhich‘external’–althoughitwasintendedto
beapparent,notlabellingtheactualsolutionsasactuallybeing‘theinternal
solution’mayhavecauseparticipantstobeunsureofhowtheframeapplied.
Whatthesedifferentconditionsreallypointtoistheneedforintegrationin
orderforframestobeeffective.InthecaseoftheQuestionFrame,thecondition
tookacoupleofquestionsfortheinformationtobeintegratedandasaresulta
coupleofquestionsfortheframingtobepredictive.IntheOver‐Frametheframe
neverbecameintegratedforwhateverreason,althoughitwasknownbythe
participantssoithadtheeffectofconfusingthedecisionmakingprocessasa
resultofalackofintegration.
Integration
Thisstudyprovidesfurtherevidencethatthetraditionalunderstandingof
framingeffectsandboundedrationalityisinsufficienttoexplainactivityinmore
dynamicenvironments.Thatshouldnotbetakentosaythatpreviousresearchis
notuseful;clearlytheAsianfluframingutilisedinthisstudy,andtheseriesof
framingmechanismsemployedinthepreviousexamplearedrawnfrompre‐
108
existingliteratureanddemonstratethattheyarestilleffectiveinongoingtasks.
However,thewayinwhichframesworkinthisdifferentcontextclearlyhasa
numberofuniquepropertiestoitthatdemandfurtherstudy.
Theemergingpattern,asnotedabove,appearstobethatintegrationiskey.That
is,thatbothataskandtheframemustbeprocessedandunderstoodinorderfor
theframetohaveaneffectonthetask.Evidenceofthiswasseenintheprevious
experimentwheretheframewasnotappliedtodiscardingwhenitcouldhave
been,thenfurtheredherewhereframingtooksometimetobeestablishedinthe
QuestionFramecondition,andalackofintegrationdisrupteddecisionmakingin
theOver‐Frameconditionatthestartoftheexperiment.Integrationisprovingto
beakeywayforexplainingifframesareeffective(orineffective).
Thismakessenseonabasiclevel–itisnecessarytounderstandinorderforitto
influenceyourdecisionmakingafterall.Howeverinalltheexamplesgiven,it
wasnotthattheinformationfailedtobeunderstoodatall.Indeed,inthecard
gamestudydiscardingwasunderstoodasamechanismandimplementedwith
onlysmallmis‐stepsattimes.Similarlyinthisexperimentparticipants
understoodtheinstructions–theydidnotaskforclarificationandexpressedno
confusionafterwards.Whatisgoingonisnotasimplefailuretoprocess,oreven
afailuretounderstandonabasiclevel,butratherrepresentsafailuretograsp
theimplicationsoftheinformationandapplyitappropriately.
Thesignificanceofthisfindinghasseveralaspects.Firstlyitactuallyidentifiesa
largepartofwhywewouldanticipateframingeffectstohaveadifferentimpact
inongoingscenarios–theyhavealotofmovingpartsandcomplexity.The
significanceofthisshouldnotbeunderestimated;ifintegrationatarelatively
highleveliskey,thebusierandmorecognitivelydemandingtheenvironment
thegreaterthepossibilitythataframecouldgetlostinthemilieu.Itmaybethat
framingfailstohaveaneffectnotbecausetheframeitselfisbadlydesignedor
incapableofaffectingsomething,butinagivencontextitsimplydoesnot
generateenoughattentiontobeeffective.
Thisraisesthequestionofhowframesareattendedto,andthereforeintegrated
intoaparticipant’sunderstanding.Fortunatelythisexperimentprovidesauseful
109
exampleofjustthiswiththeQuestionFrame.Herethesameframewasfully
integratedintothequestion,bothintermsofdescribingtheproportionsinterms
oftheproblemathand,butalsointermsofbeingpartofthetextthat
participantswereengagedwithreadingforanunderstandingoftheproblem.
Theinformationwasthere,itwasavailableanditwaspartoftheprocess
already.Thisraisesquestionsabouthowseparateaframecanbefromthe
informationitisrelevantto.
Additionallythereariseissuesinregardstopredictabilityandforecastingof
behaviour.Oftenwhenapplyingpsychologicaltheoriestopracticalsituationsit
isnecessarytohedgeprobabilitiesanyway–ifexposedtothisstimulus,70%of
peoplewillrespondinaparticularway,or60%ofparticipantsgenerally
respondfavourablytoaparticularsortoftherapy.Howeverwhatthis
experimentillustratesisthatwithframingadifferentsortofcalculationapplies.
Thequestionisnotonlyhowparticipantswillreacttoagivenframe,butalso
whethertheframewillbeintegratedandthusappliedatall.
Inthisexperimentitwasaperfectlyreasonableprediction,basedonprevious
experience,thatinsertingaframeintotheinstructionswouldbesufficientto
createanoticeableandstatisticallysignificantdifference.Thatthiswasnotthe
casewasunexpected,butinamannerthatisperfectlyconsistentwiththe
developingunderstandingoutlinedinthisthesis.However,thenatureofthat
understandingisthatitisdifficulttopredictaheadoftimeexactlywhatwillbe
integratedornot.Asetofpropertiescanprobablybeassembledtoaidwiththis
task,butgiventhateveninthisrelativelycontrolledenvironmentitwashardto
do,itislikelythatincreasinglyrealisticenvironmentswouldprovideincreasing
levelsofchallengeforthatprediction.Asthecomplexityincreasesthepossibility
thatsomeotherfactorwithinterferewithexpectedintegrationincreasesalso.
Asaresultofthis,itseemsprudenttosuggestthatthisisseenasatwo‐stage
processforsuchpredicting.Theeffectivenessofaframeshouldbeevaluatedas
onefactor,butasaseparate(althoughinteracting)factor,thelevelofintegration
shouldalsobeassessed.Inthisexperimentitwastheintegrationlevel,notthe
typeofframethatprimarilyaccountedforthedifferencesbetweenconditions,
110
giventhattheframesweremathematicallyidentical.Thedangerthisillustrates
isthatcaremustbetakennottorejectaframingapproachasineffectivedueto
theframeitself,ratherthanalackofintegration.
Integrationcanthereforebeseenasanewdimensiontobestudiedforbounded
rationality.InthesamewaythatKahnemanndescribesapartiallistofproperties
thatcanaffectavailabilityfordecisionmaking,itisnothardtoenvisionalistof
similarlydefinedpropertiesthatwouldstarttomapthelandscapeofwhat
affectsintegrationforframingeffects.Indeed,itwouldbethoroughly
unsurprisingifthesepropertiesalsoturnedouttorepresentmoregeneric
propertiesofperceptionanddecisionmaking.Whatframingcouldrepresent
here,then,wouldbethemeansbywhichintegrationcouldbestudied.Givena
consistentbiasthatisknowntoexistwithanestablishedframe,variablelevels
ofpotentialintegrationcouldbesetandtheireffectivenessgaugedbytheir
effectivenessinframing.Indeed,itwouldbeinterestingtoknowifintegration
and/orframingarebinarypropositions,orifitwouldbepossibletocreatea
partialorlesserframingeffectthroughpartialintegration.
Confidence
Thesecondimportantoverallobservationtocomeoutofthisstudyconcernsthe
confidencemeasure.Itwasbothreadilyacceptedandunderstoodby
participants,aswellasdisplayingeffects.Thesedemonstratethatitwasauseful
andrelevantfactortobemeasuring,aswellasapparentlyaccuratelycapturing
whatitwasintendedto,lendingconfidencetoitsappropriateness.Some
conclusionscanbedrawnfromitsuse.
Firstly,confidencevariedsystematicallybycondition.Thebroaderimplicationof
thisissimplythatwhenframingaconceptinordertodriveopinion,itislikely
thatconfidencewillbeimpactedatthesametime.Decisionsmaybeaffects,but
thischangewillcomewithadditionalconsequences‐peoplemaybemoreorless
enthusiasticaboutthechoicetheyaremaking,andmoreorlesswillingtotake
risksasaresultofit.Whenframingopinion,simplydeterminingthedirectionin
whichadecisionwillnowgoisnotsufficienttopredicttheoutcomeofthe
111
actionsthatwillbeundertaken.Framingwillaffectnotjustwhatisdone,but
alsohowitisdone,andtheseadditionalfactorsarejustasimportant.
Noting‘additional’factorspluralinthepreviousparagraphwasnotanerror.
Thatconfidenceisaffectedistheonlystatementwecanbeconfidentof
currently,buteverythingaboutthatresultsuggeststhatotherelementscould
havesimilarproperties.Itisreasonabletohypothesise,ashasbeendoneearlier,
thatconfidenceisnottheonlyfactorthatislikelytobeaffectedbyframing.Butit
isalsotruethatthefactorthatwedohavesolidevidencefor–‘confidence’could
itselfbeunpacked.Itwasavariablethatwasdeliberatelyleftbroadandfor
interpretationbytheparticipants,butitwouldnotbedifficulttoimaginethatit
mightbeparsedintomorespecificaspectsofthatconcept.Alternativelythe
behavioursitmightinfluencecouldbeseparatedanddifferentiallyexamined.In
theriskquestionnaireusedinthisexperiment,differenttypesofriskfactorsuch
asphysicalorfinancialwereidentifiedandavailabletoberatedseparately.With
thisdataset,parsingtheresultsfurtherbythatmetricwouldhaveresultedin
lowpowerandpotentiallyunevengroups,butfurtherworkmightfindthat
framingmorespecificallytargetsfinancialconcernsthanitdoesphysicalones,
forinstance,andthatthisaffectstheconfidenceasaresult.
Whatisalsonotableaboutconfidenceisthatitdidn’tsimplyvarybycondition,it
interactedaccordingtofeedback.Thesamepointsaboutavarietyoffactorsthat
mightaffectconfidencecouldbemadeaboutfeedbacktype,anditshouldbe
notedthatthisisfurtherevidenceofthepotentialcomplexityofframingina
dynamicenvironment.Howeverregardless,themoresignificantresultisthat
framingdidnothaveagenericraisingorloweringofconfidenceeffect.Thisis
anotherexampleoftheneedtocontextualiseanyunderstandingwhenlookingto
makepredictionsaboutresponses,orrelativeconfidence.
Ifthisunderstandingcanbecontextualisedandmaderobustenoughforagiven
situation,theunderstandingcouldbeusedtosystematicallyaffectthe
performanceofactorsinagivensetting.Fromthedataaboveweknowthat
framingcanaffectconfidence,andweknowthatresponsetimecorrelateswith
confidence–themoreconfidentyouarethefasteryoumakeadecision,andthe
112
lessconfidentthelongeryoutake.Itisnothardtoseehowthiscouldbeapplied
in,forinstance,acombatscenario.Ifaframecanbeintroducedtotheopposing
forcethroughtheuseofpropagandaorothermeansitdoesnotevenhaveto
actuallyaffectthedecisionsthatthepeoplewhoreaditmake.Theycouldmake
thesamedecisionsatthesameplaces,butdosoeverysoslightlyslower,asa
resultoftheframe.Andinacombatsituationthosecriticaldecisions–whento
move,wheretomove,whethertoopenfire–canliterallybelifeordeath
dependingonsplitsecondtiming.
ImplicationsfortheModel
Theimplicationsthisworkraisesforthemodelareessentiallysupportive.The
stepofintegration,inparticular,islookingmoreandmoreimportantasstudies
continue.Thisraisesthepossibilitythatasastepitis,infact,toosimplistic.The
informationintheinitialmodelisunderstoodsimplyasaneither/orstate,but
increasingevidenceisthatthisisnotthecase.Thingscanbeperceivedand
integratedtoadegree(suchasthediscardingrule,ortheOver‐Frameinthis
example)andnotgoontoinfluencedecisionmaking.Participantswereclearly
awareofsomethingssothereforetheymusthavebeenintegratedonsomelevel,
butnotsufficientlyforthatunderstandingtohavetheexpectedinfluence.
Recognitionofthecomplexityofthisstepwillbenecessary,some
discriminatorymechanismfordetermininghowandwhydifferentelementsare
integratedtoadifferentdegree.
NewQuestions
Thisexperimentraisesanumberofquestionstobeaddressed,aswellas
openinguppotentialavenuesoffutureresearch,someofwhicharetoowide‐
rangingtobeaddressedbythisthesisatthispoint.
Themostbasicquestiontobetakenforwardsisoneofintegrationandfuture
influence.Inthisexperimentmuchhasbeenmadeofthedifferencein
integrationbetweenthetwoattemptedframings,withthesuggestionthatthisis
trueataplanninglevel,withtheframehavingbeenintegratedintodecision
makingbiases.Ifthisistrue,thereshouldbeaneffect(oreffects)seenasaresult
ofremovingtheframeandposingthesamesortsofquestions.Ifthereisreally
113
integration,lingeringframeeffectsshouldbeseen.Additionally,todatethe
framingeffecthasbeenestablishedonlywhenpresentedaspartoftheproblem
tobesolved.Giventhatthisisunlikelytobethecaseintherealworld,ifthe
framecanbepresentedasasecondarysourceofevidenceandbeintegratedinto
thedecisionmakingstilltherewillbegreatercauseforconfidenceatthe
effectivenessofutilisingthisunderstandinginrealworldsituations.
Secondlytherearefurtherquestionsofapplicabilityandethnographicvalidityto
address.Betweenthecardgameandthisstudy,thephenomenahavebeen
robustlyestablished,andstepscanbetakentoapplytheunderstandingtoan
ongoingareaofresearch.
114
CHAPTERSIX–FramePersistenceandCyberInfluenceasaFramingMechanism:TheHerbalStudy
Introduction
Previouschaptershaveexaminedtheeffectofframingonstrategicbehaviourin
tasks.Thischapterwillre‐utilisethepreviouslyemployedseriesofforced
choicesandconfidencemeasures,whilstemployingadifferentframing
mechanismemployedinamoreecologicallyvalidmannerandexploring
whethereffectspersistoncetheframeisremoved.
Setting
Thesamebasicparadigmwasemployedasthepreviousexperiment,withthe
significantdifferencethattheframingmechanismwasremovedfromdirect
associationwiththeproblemdescription.Asocialmediafeedwaschosentofulfil
thisrole,beingbotheasilyseparablefromaproblemdescriptionandasourceof
potentialinfluencewithsignificanceintherealworld.
Itisacommonassumptionthatsocialmediaplaysasignificantroleindriving
behaviour.Examplesofthisincluderelativelyinnocuousactivitiessuchasthe
coordinatedmass‐buyingofasongtogetittonumberoneintheUK(News,
2009),andmoresignificantglobaleventssuchasthe‘ArabSpring’seriesof
uprisingsandunrestinthemiddleeast(Howard,2011;Saletan,2011)andthe
UKriotsof2011(Halliday,2011).
Researchhasnotaddresstheideathatsuchsitescanbeframing,buthave
concludedtheyareinfluentialtosomedegree.Studieshaveexploredhowusers
shareandpromotephysicalexerciseviathesite(Kendall,Hartzler,Klasnja,&
Pratt,2011),howacademicsciteonit(Priem&Costello,2010),andevenhow
Governmentsuseit(Wigand,2010).Therehavealsobeeneffortstounderstand
themechanismsdrivingsuchinteractionandtheireffects.Studieshavelookedat
usingittoexaminecurrentattitudesandbeliefs(Marshall&Shipman,2011),
andquestionshavebeenpersistentlyraisedaboutwhethertheinformation
115
beingsharedisaccurateornot(Castillo,Mendoza,&Poblete,2011;Mendoza,
Poblete,&Castillo,2010)andestablishedthatinfluencewithintwitteris
distributedandnotlimitedto‘authoritative’sources(Bakshy,Hofman,Mason,&
Watts,2011).Forthesereasons,socialmediapresentsareal‐worldexampleofa
sourceofinformationthatispotentiallyinfluentialandcouldplausiblybe
framing.
Design
Theexperimentconsistedoftwoparts:aseriesofquestionnairesanda
computerprogramthatparticipantscompleted.
Therewerefourquestionnaires,threeofwhichwereadministeredpriortothe
experiment,andonethatafterwards.Thethreebeforehandconsistedoflikert
scaledquestionsthatmeasuredattitudestoalternativemedicines(seeappendix
twenty),introversion/extraversionandrisktaking(Francisetal.,2006;Patton
etal.,1995)(seeappendicessevenandnine).TheI/Eandrisksheetswerethe
samemeasuresusedinthepreviousstudystandardisedfrompriorresearch.
Riskandattitudetoalternativemedicinesweremeasuredaspotentiallybiasing
factors.Introversion/extraversionwasmeasuredasabaseline:itwasnot
expectedthatitwouldhaveaneffectonthetask.Thealternativemedicinesheet
wasconstructedspecificallyforthisworkastherewerenostandardmeasures
available,andwasintendedasaroughguide.Itwasadministeredonthe
assumptionthatpre‐existingbiasestowardsthesubjectmatterwouldexist,
althoughtheexperimentwasdesignedsuchthatthisshouldnothavebeena
factor.Afinalquestionnairecoveringuse‐ofandattitudestowardssocialmedia
wasadministeredattheendoftheexperiment(seeappendixtwenty‐one).
Method
Thetaskconsistedofaseriesofthematicallylinkedbinarychoices,pairedwitha
confidencescale.Therewere18questionsintotal,dividedintothreesections:an
initialelevenquestions,afinalfiveandthentwo‘rethink’repeatedquestions.
Thefirstsectionwouldconsistoftheframedquestions(apartfromintheneutral
condition),withtheframingeffectachievedbyamocked‐upTwitterfeed.There
wouldthenbean‘event’;aframednon‐responseeventfollowedbythefinalfive
116
questions,wheretherewouldbenoframing.Afterthis,twoquestionswouldbe
re‐presentedtotheparticipants.
Foreachchoice,thecomputerscreenwouldshowaparticipantthreegrey
panels,withabuttonbeneatheachlabelledas‘showtweets’‘showdecision’and
‘makechoice’respectively(seefigures7.1‐7.3).Participantsbeganthesectionby
pressingthestartbuttontotheleft,whichwouldfirstdisplaythetweets.
Participantsthenhadtoviewthedecisionnextbyclickingonitsbutton,atwhich
pointasitwasrevealedthetweetswerere‐concealed.Afterthistheycould
proceedtothedecisionmakingsection,whichconsistedoftwobuttons(onefor
eachcompany)andasevenpointLikertscalemeasuringconfidence.Participants
werefreetoviewwhicheversectiontheywishedforaslongastheyliked,and
returntothemasmanytimesastheywished,butatanyonetimeonlyone
sectionwasvisible.Everytimeanybuttonwaspressed,theprogramrecorded
timing,choiceandotherrelevantdata.
117
Picture6.1:Screenshotofanexampletestscreen,withTwitterfeedcurrently
visible
Picture6.2:Screenshotofanexampletestscreen,withthescenariotextvisible
Picture6.3:screenshotofanexampletestscreen,withchoicebuttonsvisible
118
Thesettingofthetaskplacedtheparticipantintheroleofabuyerofalternative
medicineproducts(specificallyherbalremedies)foracooperativegroupof
shops.Foreachdecision,participantswererequiredtochoosebetweentwo
fictionalcompaniesthattheycouldorderfrom.ThesewereQuetia,locatedinan
Antipodeannationandrepresentingslightlyhighercostbutgreaterqualityand
Astor,locatedinAsiaandrepresentingslightlycheaperproductsandgreater
authenticity(seeappendixtwenty‐twofortheexperimentalinstructions).These
distinctionswereintentionallygivenasmildandwerenotintendedtobe
biasing,butrathertoaddagreatersenseofrealismtothetask.Theseriesof
decisionsthroughtheexperimentdescribedbothaseriesofsmall,unconnected
dilemmasandchoices,butalsoanoverarchingnarrativeaboutacontaminated
batchofaproductthatbothcompaniesproducecausingsickness(seeappendix
twenty‐fourforanexampledecision).Theeventdescribedtheeventualoutbreak
ofwidespreadproblemsassociatedwiththisandwasframedinthemannerof
theAsianfluexample(seeappendixthirty).Theremainingfivequestionsthen
dealtwithrepercussions,differingreactionstotheproblemandcleanup.
Progressionthroughthequestionswaslinear:allparticipantsreceivedthesame
questionsinthesameorder.Theyweretold,however,thattheirresponsesand
decisionsaffectedwhathappenedandthequestionsthatwereaskedasaresult.
Thiswasdoneintentionallytomotivateparticipantstothinkofthetaskasa
singletaskratherthanaseriesofindividualdecisions.
Feedbackwasprovidedaftereachchoicewasmadeaccordingtotwometrics,
retailerconfidenceandprofit.Bothweredescribedqualitativelyasvariationsof
‘small/medium/large’lossorgain.Theexactcombinationsofthesemetrics
variedbetweenquestionsinordertogivetheillusionofrandomchance,which
participantsweretoldbeforehandexistedbutinrealitytherewasnone.Both
conditionswerebalancedsuchthatifaparticipantchoseonlyonecompanyor
theother,theywouldevenouttohavethesameresult(thiswastrueforboth
setsofquestions,beforeandaftertheevent).Similarly,theactualfeedback
variedwithwhichchoicewas‘better’onalinearbasis,A‐Q‐A‐Qetc.The
119
exceptionstothiswerethefirstandlastquestions,whichrewardedequally
regardlessofparticipantchoices.Thereasonforthiswasthatwithboth
questions,itwasdesiredthatparticipantswouldfeeltheyhadmadethe‘right’
choicewhichevertheypicked.Particularlyforthefirstquestionitwasfeltthat
participantsshouldnotbegintheexperimentbeingpunishedforachoice.The
effectofthiswasthataparticipantthattheoreticallychoseallofonecompany
wouldseetheexactsamefeedbackoverallthataparticipantwhochosetheother
companywould(seeappendixtwenty‐threefordetails).
Framingwasachievedviaamocked‐uptwitterfeed.Itwaschosentousetwitter
forseveralreasons.Firstlyitisarecentandwell‐coveredwebsite,ensuringboth
generalfamiliarityevenamongnon‐usersaswellasreflectingarecenttrendin
onlinesocialnetworkingtowardsbroadcasting.Secondlyitiscommontosimply
‘follow’peopleontwitterforinformationwithoutbeingfriendswiththemand
knowingthempersonally(asisoftenthecasewithfacebook).Thismakesit
moreappropriateforthetaskof‘informationgathering’.Thirdly,therelative
simplicityofitsinterfaceanddisplaymadeiteasiertoincorporateintoan
experimentwithoutseemingtotallyunnatural.
Theframeitselfwasconstructedalongthelinesofheuristicelicitation,as
describedbyKahnemanandTversky(Kahneman&Tversky,1973).Inthat
experiment,apersonalitydescriptionofdubiousqualitywasshowntobegiven
significantweightindeterminingprobablefieldofstudyforastudent.Inthis
experiment,aninformationofdubioussourceandveracity(thetwitterfeed)was
tobeusedtoguideparticipant’sdecisions.Participantswereinformedthatthe
feedrepresentedpeoplebeingfollowed,butthattheyhadnogreaterauthority
thanthat.
Eachquestionwasaccompaniedbyanon‐interactiveimageofatwitterfeed(see
appendicestwenty‐fivethroughtwenty‐sevenforexamples).Allimageswere
generatedbycreatingthetwitteraccountsinquestion,givingthemnames,
genericprofilepicturesandthenhavingthem‘tweet’theappropriate
informationinagivenorderbeforetakingascreenshot.Foreachquestion,the
twitterfeedshowedfivetweets,fromthesamefivetwitteraccounts.
120
Theinformationwithinthesetweetswasdeterminedsystematically.Neutral
conditiontweetsweredeliberatelychosentohavenothingtodowiththe
experimentoranyaspectofit.Occasionallyatweetwouldcommentuponthe
informationfromthedecision,butnevertogiveanopinionoranyactual
informationaboutit.Forinstancewhenthedecisionrelatedtothefactthat
customerswerebuyinglessherbalmedicine,oneofthecommentswas
‘EveryoneseemswaryofHerbalmedicinethesedays’–asimplere‐statingofthe
decisiontocome,andnotsomethingthatshouldaffectthedecision.Theyalso
containedthenewstweets,asdescribedbelow.
Thefivetweetsthatappearedinaframingtwitterfeedconsistedof:
1) Anewstweet,statingthedecisionasnewsandprovidinga(non‐
functional)linktoastory.
2) APro‐frametweet,restatingthepro‐rationalemadewithinthe
description.
3) AnAnti‐frametweet,restatingtheanti‐rationalemadewithinthe
description.
4) Asecond,novel,pro‐frameopiniontweet
5) Atweetgivingastatementofpersonalpro‐framebehavioureither
enactedoranticipated.
Themainproandantiframetweetswereidenticalbetweenconditions,although
theyswitchedrolesdependingwhichsidewasbeingframed.Thenewstweet
alsostayedidentical,meaningthattheonlyinformationthatactuallychanged
betweenthetwoconditionswasthesecondaryoption,andstatementofaction.
Intheneutralcondition,nobiasinginformationwasincludedeitherway.
Theorderthattheaccountsappearedinthefeedwasrandomized,andthenkept
constantforeachquestionbetweentheconditions.Theorderinwhichtheframe
informationappeared(andthuswhichaccountwassayingit)wasalso
randomisedandkeptconstantbetweenconditions,meaningthateachcondition
wasseeingthesameorderofproandantiframetweets.Thenewstweet,
naturally,wasthereforeinthesamepositionbetweenallconditions.
121
Hypotheses
Firstlyitwaspredictedthatframingwouldbeseeninthisexperiment,thatthe
twitterfeedwouldsuccessfullybiasparticipantresponsesintheintended
direction.
Secondlyitwaspredictedthatframingeffectswouldbecontinuetobeseenonce
theframewasremoved,aswouldbeconsistentwiththeintegrationtheory
proposedinthepreviousexperimentwhichproposedthataframebecame
integratedintodecisionmakingprocesses.
Thirdlyitwaspredictedthatsystematicdifferenceswouldbeshownin
confidence,andothersourcesofdata,betweenframedandunframed
participants.Itwaspredictedthatframedparticipantswouldbemoreconfident
thantheirunframedcounterpartswhenmakingdecisions.
Finallyitwaspredictedthatthenon‐response‘event’wouldframesubsequent
choices,alsoconsistentwiththeintegrationtheory.
Results
Theresultsfromthisexperimentaretakenfromanumberofmeasures,
primarilythechoicesmade,theconfidenceinthosechoicesandtimingdata.
PretestingData
Allparticipantsweregivensurveystocompleteasdetailedabovemeasuring
theirattitudestowardsalternativemedicines,risktakingpropensity,familiarity
withsocialmediaanddegreeofintroversion/extraversionaswellasgathering
generaldemographicinformation.
48participantsintotalweretested,dividedequallyintotherelevantconditions.
Therewereanequalnumberofmaleandfemaleparticipantsineachcondition
andtheexperimentasawhole.Theageofparticipantsrangedbetween19and
46,withameanageof27,andmedianof26.Mostparticipantswereinsomeway
involvedwiththeuniversity,althoughtherewasamixofundergraduateand
122
postgraduates.Therewerenosignificantdifferencesbetweenanyofthegroups
intermsofagecomposition.
Asubsetofthe48participantswasconductedatdistanceduetodifficultywith
obtainingparticipantsin‐person.Thisinvolvedparticipantsrunningthe
computerprogramonaPCattheirlocation,andfillingouttherelevant
questionnaireformsinorder.Itwasensuredthatallparticipantswereusingthe
sameequipmentasthosetestedinperson,andwhilstthetestwasbeing
administeredtheywereobservedviaone‐wayvideoconferencingsoftwareto
ensurethattheydidnottakeabreak,drink,eatorengageinanyotheractivities
thatwouldhavebiasedtheirresults.Thissubsethadameanageslightlybelow
theaverage,havingamedianandmeanageof24.Theparticipantswereevenly
distributedbetweenthe6conditionswithoneineach,andtheirresultswere
collectedandcomparedtothein‐persongroupresults.Nosignificantdifferences
werefoundinthetimetakenfortasks,decisionsmadeorconfidence,andasa
resultitwasconcludedthatconductingtheexperimentatdistancehadno
significanteffectonthetask.
Thedifferentgroupsweretestedbetweenforthesurvey‐measuredtraits.No
significantdifferenceswerefoundbetweengroupsforanyofthemeasures
obtained.Itwasconcludedthatpreexistingbiaseswithinthegroupswouldnot
biasanyoftheresultsobtained.
Theneutralconditionparticipants’answerstothefirst11questionswerethen
usedtoseeiftherewasacorrelationbetweenanyofthefactorsandthechoices
made.Thiswasdonebecausethetraitscouldstillbepredictive,evenifthey
wereevenlydistributedandthusnotbiasingtotheexperimentaldesign.The
neutralconditionalonewasused,becausetheframedconditionswould,bytheir
verynature,bebiasingoverandbeyondthatwhichcouldbepredictedbythe
traits.Nocorrelationwasfoundbetweenthenumberoftimesaparticipant
choseAstorandeitherattitudetowardsalternativemedicine,or
introversion/extraversion.Acorrelationwasfound,however,betweenrisk
takingandAstorchoicesusingSpearman’sRhor(16)=.567,p=.022.This
123
indicatedthatparticipantssawQuetiaasthe‘lessrisky’choicebetweenthetwo
companies,somethingthatwillbereferencedlater.
Choices
Forthepurposesofthissection,the‘choice’inquestionreferstowhichcompany
waschosentobuyfromforeachquestion.
Totalswereamassedforthefirstelevenquestionsineachcondition:Framedfor
Astor,framedforQuetia,(subsequentlyreferredtoaseitherAorQframes)and
Neutral.Eventconditionswereignoredasafactor,sinceuntilthatpointall
conditionswereidentical,andframingonlyoccurredupuntiltheevent.
Itwasfoundthatparticipantswereframedinanapproximately7:3ratio
(framedchoicetounframedchoice)inbothdirections.Intheneutralcondition,
participantsweresplitapproximately50‐50betweenthetwochoices.
Figure6.1:ParticipantChoices(A/Q)(yaxis)bycondition(A/N/Q)(xaxis)forthe
firstelevenquestions
Aseriesofchisquarecalculationswereperformedonthedata,anditwasfound
thatboththeAframe(c2(1,N=352)=13.78,p=.000)andQframe(c2(1,N=
352)=19.99,p=.000)choicesweresignificantlydifferenttotheneutral
condition(andthustoeachother).Thissuggeststhattheframingwassuccessful
124
ininfluencingthedecisionmadeandthattheframewasapproximatelyequalin
powerinbothdirections.
Thisdatawasalsovisualizedbypresentingthedatabyquestion:
Figure6.2:PercentageparticipantchoiceofA(yaxis)byquestionnumber(xaxis)
forfirstelevenquestionsbycondition(A/N/Q)
Thesamecollatingofdatawasperformedforthelast5(unframed)questions.
Firstthedatawasdividedaccordingtotheframeapplied.
Figure6.3:Participantchoices(A/Q)(yaxis)bycondition(A/N/Q)(Xaxis)forthe
finalfivequestions
125
ComparingeithertheAorQframeswiththeneutralconditiondoesnotproduce
asignificantresult,howeverdirectlycomparingtheAandQframesissignificant
(c2(1,N=352)=4.97,p=.026).Participantswerealsocomparedwhensplitby
theeventframe(QuetiaframedorAstorframed),howevernosignificant
differencewasfound.
Thedatawasalsochartedaccordingtothedifferencebetweenthefirst11and
lastfivequestionstoshowrelativechangetowardsQuetiabetweenconditions.
Figure6.4:PercentageDifferenceinchoicesmadebetweenthefirstelevenandlast
fivequestions(yaxis)byframe(xaxis),shownaspercentagechangetowards
Quetia.
TheproportionalchoiceofAwasthengraphedforthelastfivequestions(Figure
6.5below).
126
Firgure6.5:ParticipantchoiceofA(proportion)(yaxis)byquestionnumber(x
axis)forfinal5questionsbycondition(A/N/Q)
Finally,thesameanalysiswasappliedspecificallytothelastquestionofthemain
experiment,question16.Thiswasdonebecausethatparticularquestionasked
Figure6.6:Participantchoices(A/Q)(yaxis)bycondition(A/N/Q)(xaxis)forthe
finalquestion
participantstomakeafinalchoiceofwhichofthecompaniestheywouldchoose
tobuyoffforayeargoingforwards.Consequently,itcanbeseenasa‘final
preference’.
127
Itwasfoundthatalthoughtherewasnosignificantdifferencebetweenthe
Neutralconditionandeitherframe,therewasasignificantdifferenceattheone‐
tailedlevelbetweentheAandQframesusingFisher’sExactTest(p=.041).The
lownumberofresponsesmadeFisher’sExactTestmoreappropriatethanaChi
Squareinthiscircumstance.
Confidence
Theoverallconfidencescoreresultspreandpostevent(firstelevenverseslast
five)werecomparedforeachcondition.
Figure6.7:Averageconfidence(yaxis)bycondition(A/N/Q)andquestionset(x
axis).Errorbarsshowstandarderror.
A3x2ANOVAwasperformedonthedatawithFrame(A,N,Q)andquestionset
(First11,last5)asbetweensubjectfactors.Maineffectswerefoundofquestion
set,F(1,762)=7.10,p=.008,andframe,F(2,762)=3.73,p=.024.Thesewere
qualifiedbyaninteraction,F(2,762)=3.75,p=0.24.AseriesofMann‐WhitneyU
testsfoundthatinthefirst11questionstherewasnosignificantdifference
betweentheAandNframeconditions,butthatQframedconfidencewas
significantlydifferenttobothAframed,U=13037,p=.008,Z=‐2.65,and
Neutralconditions,U=13397,p=.03,Z=‐2.17.Betweenquestionsets,onlythe
Aframewassignificantlydifferent,U=5203,p=.001,Z=‐3.45.Inthelast5
questions,BoththeA‐frame,U=2552,p=0.24,Z=‐2.26,andQ‐Frame,U=
128
2637,p=0.48,Z=‐1.98,weresignificantlydifferenttotheNeutralcondition,and
identicaltoeachother.
Choice
Astor Quetia
A Frame 5.00 4.79
Neutral 4.83 4.79 Condition
Q Frame 4.20 4.67
Table6.1:MeanConfidenceforparticipantchoicesbyconditioninfirsteleven
questions.Allfiguresaccurateto3sf
Confidencewasalsoexaminewithinchoicesmadeforthefirst11questions
(table7.1above)Ascanbeseenfromtheabovetable,confidenceinQuetiadoes
notvarygreatly(orsignificantly),whetherframedorunframed.Confidence
does,however,varysignificantly,U=1923,p=.000,Z=‐3.878,inAstorwhen
framed.
Thiswasrepeatedforthelastfivequestions:
Choice
A Q
A 4.26 4.27
N 4.43 4.92 Condition
Q 4.22 4.35
Table6.2:MeanConfidenceforparticipantchoicesbyconditioninfinalfive
questions.Allfiguresaccurateto3sf
Nosignificantdifferenceswerefoundbetweenchoicesforthesequestions.Since
thevarianceappearstobemainlyintheAstorchoicesthathypothesiswastested
bygraphingtheaboveinformationforeachchoicetypebetweenquestionsets.
129
Figure6.8:Confidencescores(yaxis)forAchoicessortedbyframe(A/N/Q)and
questionset(xaxis).ErrorBarsshowstandarderror.
Figure6.9:Confidencescore(yaxis)forQchoicessortedbyframe(A/N/Q)and
Questionset(xaxis).ErrorBarsshowstandarderror.
ThesechartsappeartoshowdivergenceinthefirstelevenquestionsforA
values,thenconversanceinthelastfive.ThepatternisreversedforQvalues,
whichareroughlyequalforthefirst11questions,thendivergeforthelastfive.
Thedatawasthensortedaccordingtoconfidencewhenachoiceismadethat
goesalongwiththeframe(pro‐frame)andagainstit(anti‐frame).
130
Figure6.10:Meanframedparticipantconfidence(yaxis),organizedbyquestion
set(xaxis)andframerelativechoice(Pro/Anti).Errorbarsshowstandarderror.
Thereisasignificantdifferencebetweentheconfidenceparticipantshavewhen
makingthechoicetheyhavebeenframedtowardsoragainst,U=10705,p=
.019,Z=‐2.351.Bothgroupsfallinconfidenceonceframingisremoved,butonly
thepro‐framefallissignificant,U=9391,p=.004,Z=‐2.873.
Confidencewasthenchartedbyquestion.
Figure6.11:MeanParticipantConfidence(yaxis)byQuestionnumber(xaxis)for
eachcondition(A/N/Q)
131
Althoughthedataissomewhatnoisy,hereweseethattheconfidenceoftheA
andQframesalmostmirroreachother,andappearstomatchupwiththe
patternofchoicesbyquestionasseenintheprevioussectiontoadegree(Figure
6.2).
Therelativelackofnoisewhenunframedcanalsobeseeninthelastfive
questions.
Figure6.12:MeanConfidencescores(yaxis),byframe(A/N/Q),forthelastfive
questions(xaxis)
TimingData
Timingdatawasobtainedforallparticipantsaspartofthecomputerprogram.
Timeswerecapturedatanypointthataparticipantclickedonabuttonormade
achangeinsomeothermannerthatcausedtheprogramtoact.Theexperiment
wasdesignedtoisolatetheconstituentpartsofathedecisionmakingprocess–
Tweets,Descriptionandchoicebutton–sothatonlyonewasviewableatatime,
creatingadirectmeasureofhowmuchtimewasspentwitheach.Itisassumed
thattimespentwithagivensectionisanindicatorofattentionandcognitive
resourcesbeingused.
Althoughparticipantscouldgobackandforthbetweensections,thisanalysis
createsacumulativevalueofthetotaltimeforthepurposesofanalysis.The
numberoftimesasectionisviewedwasrecordedseparately.
132
Abaselinereadingspeedwasobtainedbyusingtimingdatafromthe
instructionspage.Althoughnotsimplyarawmeasureofreadingspeed(asthere
wasobviouslyacomprehensionaspecttothetask),itwasconsistentforall
participants.Forthedecisionsdata,wordcountswereobtainedforeach
questionandusedtocontrolfordifferencescreatedbyvaryingtextamounts.
Wordcountdifferenceswerenotaccountedforinthetweetssection,fortwo
mainreasons.Firstly,the140characterlimitprovidedlimitedpotential
variabilitybetweenconditions,andwordcountsweremaderoughlyequal.
Secondly,controllingfortextamountmakestheassumptionthatparticipants
willreadallthetext–reasonableforthedecisiontext,butinappropriatefor
twitter.Morethanafewparticipantsremarkedthatatonepointoranotherthey
wereeitherskimmingorignoringthetweets,andthenatureofatwitterfeedis
suchthatparticipantsmightwellchoosetoonlywanttoreadcontributionsfrom
oneperson,orconverselyignoreanother.Becausebothofthesefactors
contradicttheassumptiononwhichword‐countweightingwouldbeundertaken
itwasnotapplied.
ChoiceSelection
Thetimingdatainthissectionwasthetimespentphysicallyclickingthebuttons
thatindicatedthedecisionthattheyhadmade.Consequentlytherewasnotext
orotherinfluencinginformationtobeconsidered:foreachandeveryquestion
thesectionwasidentical.Therotetask‐basednatureofthissectionofthetask
providesausefulcomparisonpointwiththeinformation‐processing
requirementsseenintheothertwosections.Thedatawasskewed,asifoftenthe
casewithtimingdata,andwaslogtransformedtoaccountforthis.
A3x2ANOVAwasperformedonthedatawithFrame(A,N,Q)andquestionset
(First11,last5)asbetweensubjectfactors.Amaineffectwasfoundforquestion
set,F(1,762)=23.492,p=.000only.AseriesofMann‐WhitneyUtests
confirmedthattheonlysignificantdifferencewasbetweenquestionsets,U=
48494.500,p=.000,Z=‐5.217.Thereisaslightcorrelationbetweenquestion
numberandtimetakenr(766)=‐.340,p=.000.Chartingthedatasuggeststhat
133
thiseffectisdowntotaskfamiliarization.Removingthefirstthreequestions
removesanystatisticallydetectableeffect,supportingthisconclusion.
Figure6.13:Totaltimeinmsthatthe‘inputchoice’sectionisopen(yaxis)by
questionnumber(xaxis).
Thereisaslightcorrelationbetweenconfidenceandtimetakenr(766)=‐.186,
p=.000(seeFirgure6.13below)suggestingthatcertaintyaboutachoicepredicts
howquicklyitismade,aswouldbeexpected.
134
Figure6.14:Meantime(logtransform)takentomakeachoice(yaxis)by
confidencerating(xaxis)forallparticipants.
DecisionTextSection
Thetimingdataforreadingthedecisiontextwastreatedasdescribedabove.It
wasadditionallyconvertedtoitslogarithmicvalueinordertofulfillthe
requirementsofparametricanalysis.
Comparingtheamountoftimetakentoreadthedecisiontextbetween
conditions,itwasfoundthattherewasnodifferencebetweentheframed
conditions(AandQ).A2x2ANOVAwasperformedonthedatawithFrame
Status(Framed/Unframed)andquestionset(First11,last5)asbetweensubject
factors.Amaineffectwasfoundforframestatus,F(1,762)=18.015,p=.000
only.However,bothquestionset(F(1,762)=3.164,p=.076)andaninteraction
(F(1,762)=3.342,p=.064)wereclosetosignificance,andthetestfailed
Laverne’stestofhomogeneity.
Examiningfrequencydistributionsforthisdata,itwasobservedthatAlthough
theoveralldatawasdistributedinarelativelynormalfashion,individual
componentsofthedatasetweremoreerratic.Datawasstillinagenerally
normalpattern,buthadmultiplepeaksatthesametime.Thisaccountedforthe
failureofthehomogeneitytest,butalsoindicatedthatthenear‐positiveresults
meritedfurtherinvestigationastheproblemswerecharacteristicofadataset
thatcouldhavebeennormallydistributedwithmoredatapoints.
135
Figure6.15:Logofmeanadjustedtime(xaxis,0.04increments)frequency(yaxis)
bycondition(A,N,Q)forfirst11questions
Figure6.16:Logofmeanadjustedtime(xaxis,0.04increments)frequency(yaxis)
bycondition(A,N,Q)forlast5questions.
Consequently,aseriesofMann‐WhitneyUtestswereperformedonthedifferent
datapermutations.Significantdifferenceswerefoundbetweenframedand
unframedconditionsoverall(U=55529,p=.001,Z=‐3.453),inthefirst11
questions(U=27607,p=.042,Z=‐2.039)andinthelastfivequestions(U=
136
4772,p=.001,Z=‐3.211).Significantdifferenceswerealsofoundbetween
questionsetsoverall(U=56592,p=.018,Z=‐2.375)andbetweenquestionsets
intheneutralconditionalone(U=5572,p=.008,Z=‐2.673).However,no
significancedifferencewasfoundbetweenquestionsetsintheframedcondition.
Nodifferenceswerefoundbetweenthedifferentframeconditions(AandQ),and
thereforeframedresultsweretreatedasasinglecondition.
Figure6.17:MeanValuesoflogAdjustedtime(yaxis)byquestionset(First11,
Last5)(xaxis)andCondition(Framed,Neutral).Errorbarsshowstandarderror.
Figure6.18:Logofmeanadjustedtime(yaxis)takenforeachcondition(A,N,Q)
toreadthedecisiontextforeachquestion(xaxis).
137
TweetTimes
TimingdatafortheTweetviewingwastreatedinthesamemannerasdescribed
atthestartofthissection.Itwasalsoconvertedtologarithmicforminorderto
beabletomakeuseofparametricstatistics.
AswasfoundintheDecisiontimingsection,thedatawasbothinneedofbeing
transformed,andyetcorrectingtheskewdidnotcorrectthelackofhomogeneity
ofvariance.
Figure6.19:Frequencydistributionofrawtimedatainms(xaxis)bycondition(A,
N,Q).Frequencytakenin2500msbins
138
Figure6.20:Frequencydistributionoflogcorrectedandreadingspeedadjusted
timingdata(xaxis),bycondition(A,N,Q).Frequencytakenin.05bins.
Therefore,asbeforeanANOVAwasperformedonthisdatawhichwasthen
checkedandsubstantiatedbyaseriesofMann‐WhitneyUtests.A3x2ANOVA
wasperformedonthedatawithFrame(Astor‐framed,Quetia‐Framedor
Neutral)andquestionset(First11,last5)asbetweensubjectfactors.Amain
effectwasfoundforframe,F(2,762)=50.767,p=.000andquestionset,F(1,
762)=135.029,p=.000.Therewasnointeractiondetected,andthetestfailed
Laverne’stestofhomogeneity.
AseriesofMannWhitneyUtestsconfirmedthefindingsfromtheANOVA.
Overalldifferenceswerefoundbetweenquestionsets(U=33034,P=.000,Z=‐
10.642)andalsobetweenframedandunframeddata(U=38588,p=.000,Z=‐
9.298).Nodifferenceswerefoundbetweenthedifferenttypesofframe(A/Q)at
anypointsothedatasetswerecombinedforthepurposesoftheseanalyses.
Significantdifferenceswerealsofoundbetweenquestionsetsforframed(U=
11500,p=.000,Z=‐10.737)andneutral(U=4639,p=.000,Z=‐4.372)
conditions.Similarly,differenceswerefoundbetweentheconditionsinboththe
first11(U=15603,p=.000,Z=‐9.302)andlast5(U=4370,p=.000,Z=‐
4.004)questionsets.
139
Figure6.21:Meanlogadjustedtime(yaxis)byquestionset(xaxis)and
experimentalcondition(Framed,Neutral).Errorbarsshowstandarderror.
Eliminatingquestions1‐4asacclimatizingtrials,thereisstillasignificant
differencebetweentimespentwiththetweetsbetweenthefirst11andlastfive
sections(U=3164,p=.000,Z=‐3.782).Thissuggeststhatthedropisnota
statisticalartifactofaninitiallearningcurve.
Thisdatacanalsobeviewedby‐question,asbelow.
Figure6.22:Mean(adjusted)timespentonthetweets(yaxis)foreachquestion(x
axis),bycondition(Framed,Neutral)
Effectscanalsobeobservedinthenumberoftimesthatparticipantsviewthe
tweetsectionofthetest.
AscanbeseenfromFirgure6.20below,overthecourseoftheexperiment,
participantsintheframedconditionsconsistentlyreturntothetweetdatamore
thanparticipantsintheneutralcondition(U=52958,p=.000,Z=‐5.302).There
isnostatisticaldifferenceinthetwoframedconditions.Thisdifferenceholds
truewithinthefirst11questions(U=26054,p=.000,Z=‐3.568)andthelast
five(U=4696,p=.000,Z=‐4.327).Withinconditions,theframedparticipants
donovarybetweenquestionsets,whilsttheneutralparticipants’valuesfall
significantlyforthelastfivequestions(U=6032,p=.008,Z=‐2.665).
140
Figure6.23:Meannumberofviews(yaxis)ofthetweetdatabyquestion(xaxis),
bycondition(A/N/Q)
RethinkData
Asthelastpartoftheexperiment,participantsweregiventwoquestionsfrom
theexperimenttoreconsider.Theyweretoldthattheywerebeinggivena
randomquestion,althoughinactualitythequestionswerethesameforall
participantsandconditions;questions3and4.Allinformationgivenwasthe
same,exceptthattherewerenotweets,andthusnoframingdeviceofanysort
thistimearound.Thisdatawasthenanalyzedbycomparingittotheoriginal
questiondata.Becausethedatasetwassmall,onlyalimitednumberoftests
wereperformed.
Decisions
Thechoicesmadeoriginallywerecomparedwiththosemadethesecondtime
around,asdemonstratedbelow.Nosignificantdifferencewasfoundbetweenthe
twoconditions.
141
Figure6.24:ParticipantChoiceofCompany(A/Q)(yaxis)inOriginalandRethink
conditions(xaxis)
Thedatawasthensplitaccordingtotheframingthatwasusedfortheeventthat
splitthe11and5questionsections.TheconditionsinthiscasewereproQuetia
(Q+)orproAstor(A+),fororiginalandrethinktrials.Therewasnodifference
betweentheconditionsineithertheoriginalorrethinktrials.Comparing
betweentrials,therewasnodifferenceintheA+condition,howeverintheQ+
conditionbetweenoriginalandrethinktrailtherewasasignificantdifference
underFisher’sExactTestattheonetailedlevel(p=.048)
142
Figure6.25:Participantchoices(A/Q)(yaxis)inQ+conditionsfororiginaland
rethinktrials(xaxis)
ConfidenceData
ConfidenceDatawassortedandtreatedasdescribedpreviously.However,no
significantdifferenceswerefoundbetweengroups.Confidencedidnotvary
betweenoriginalandrethinktrialsoverallorwhensplitbyframe,questionor
eventtype.Italsodidnotvarywhenparticipantsthatchangedtheirmindwere
comparedagainstthosethatstayedwiththesamechoice,orwithinthosegroups
betweenthetwotrials.
However,thisnullresultdoesmeanthatparticipantconfidenceisequaltothat
intheoriginalsetofquestions‐evenfortheframedconditionswhich,as
demonstratedearliersufferedalowerconfidenceoncetheframeswereremoved
forthelastfivequestions.Thissuggeststhatthereissomepermanencetothe
decisionsastheyaremadeintermsofbelief;participantsonaveragemaintained
thesamelevelofconfidenceastheyhadbefore.
TimingData
Timingdatawastreatedasdescribedpreviously,althoughtheanalysisdidcall
fordifferentapproachesthanforconfidence.Timecouldnotbedirectly
comparedbetweentheoriginalandrethinktrialssinceparticipantsatthelatter
pointhavealreadyreadthedecisionwhenitwasoriginallypresented.Even
allowingfortheideathattheywouldreaditagain,theirperformancewouldstill
beexpectedtobequickersimplythroughfamiliarity.Instead,analysiswas
performedbetweenconditions(frame,eventetc)andthentheseresults
comparedbetweentrials.
Nosignificantdifferenceswerefoundfortimingdatarelatingtothechoice
sectionofthetask.
Thetimetakentoreadthedecisiontextwasthenanalysed.A3x1ANOVAwas
performedonthedatawithFrame(A,N,Q)asthebetweensubjectfactor.Amain
effectwasfoundforframe,F(2,92)=7.400,p=.001.AseriesofMannWhitneyU
testswerethenperformedbetweentheindividualgroups,andsignificant
143
differenceswerefoundbetweenAandQ(U=338,p=.019,Z=‐2.336)andA
andN(U=249,p=.001,Z=‐3.396).ThissuggeststhatparticipantsintheA
conditionweretakinglongertothinkabouttheirchoicethantheother
conditions.Intheoriginaltrials,nodifferencescouldbefoundbetweenanyof
theconditions.
NosignificantdifferenceswerefoundwhencomparingbetweentheA+event
andQ+eventconditions.
Extremists
Inthecourseofconductingthestudyseveralparticipantscommented
afterwards,whenaskedverbally,thattheyfelttheirresponseswerenot
appropriatefortheexperiment.Onecommentedthat“Idon’tthinkyou’llbeable
tousemydata”.Uponaskingwhattheymeantbythis,thissubsetofparticipants
commentedthattheyhadstrongunderlyingviewsaboutalternativemedicine
anditsvalue,andasaresultdidnotpaymuchattentiontoanyoftheinformation
thatmighthaveswayedthemonewayortheother.Insteadtheiranswerswere
mademainlybasedontheirpreexistingbeliefsaboutthetopic.
Theseadmissionswerenotedwhenmade.Theirdatawasthenexamined
individually,anditwasfoundthatinadditiontotheirspontaneousself‐reported
biastheyconformedtotwootherproperties:
a) Theywereinthetoporbottom5%ofrecordedscoresforattitudes
towardsalternativemedicine
b) Theirscoreswereover80%infavourofonecompanyfortheframed
portionofthetestawayfromthedirectionthatwouldbepredictedbythe
frame,orinthedirectionoftheirbiasinthecaseoftheneutralcondition.
Nootherparticipantsmetboththeserequirements,andcombinedwiththeself‐
reporting,itwasconsideredsufficienttojustifytreatingthegroupasaseparate,
emergentsetofparticipantsandexcludethemfromthemainstudy.Ofthe
group,twowerepro‐AstorbiasedandinQ‐frameconditions,andthreewere
pro‐QuetiabiasedwithtwoinaneutralconditionandoneintheAstorframe.
144
Itshouldbemadeclearatthispointthatthiswasanunexpected,emergent
groupofparticipants,andassuchwerenotcounterbalancedorcontrolledforin
anyway.Someanalysiswasperformedandispresentedbutshouldberepeated
undertighterexperimentalcontrolbeforeanyconclusionscanbedrawn.
Thattherewasaneffectatallisinteresting.Participantchoicesinthetask
reflectedtheirbeliefsinalternatemedicine:iftheybelieveditwasagoodidea,
theychoseAstor.Iftheythoughtitwasbad,theychoseQuetia.Thisisdespitethe
factthattherewasnologicalreasontomakesuchadistinction,andthepresence
ofaframe.Thispossiblyindicateslimitationstotheabilityofaframetoaffect
somepeople’sdecisionswithinparticulartopics.
Thedatawassortedbyclassingthechoicesmadebytheseparticipantsaseither
‘pro’or‘anti’theirinnatebiasratherthanleavingthemasAandQ,since
differentparticipant’sbiaswasindifferentdirections.Thesewerethenscaled
intoapercentageamountinordertocomparebetweenthefirst11andlast5
questionsets(asthesesetscontaineddifferentnumbersofdecisionstobe
made).Nosignificantdifferencewasfound.
Figure6.26:Percentageofchoices(yaxis)for(pro)andagainst(anti)inherent
beliefsinextremistsforthefirst11andlast5questions(xaxis)
Confidencewasthencomparedtothegeneralpopulation,anditwasfoundtobe
higherforextremists(U=26739,Z=‐1.970,p=.049).Therewasnodifference
145
betweenquestionsets(U=598.5,Z=‐.963,p=.336)althoughthismaybedue
tothesmallsamplesize.
Figure6.27:Meanconfidence(yaxis)fornormalparticipantsandextremistsby
questionsets(xaxis).ErrorBarsshowstandarderror.
Therearesimilarresultsforthetimingdata.Extremists,likenormalparticipants
takelesstimewiththetweetsovertime,whichcanbeseenbothasacorrelation
betweenquestionnumberandtimetakeninthefirst11questions(r(55)=‐
.4.21,p=.001)andalsowhencomparingthefirst11questionswiththelast5as
sets(U=308,Z=‐3.939,p=.000).However,theyalsotakesignificantlylesstime
thanthenormalparticipantswhencomparingbetweenbothcompletegroups(U
=25438,Z=‐2.533,p=.011).Therewerenodifferencesbetweenthenormal
conditionandtheextremistsinhowmuchtimetheyspentwiththedecisiontext.
146
Figure6.28:Adjustedtimetaketoreadtweets(yaxis)byQuestion(xaxis)for
normalparticipantsandExtremists
Withinthetimingdatathereisalsosomelimited,non‐significantindicationthat
theremaybeadifferencebetweenextremistswhenframedorunframedinthe
amountoftimetheyspendreadingtweets.Asmallsamplesizeandunbalanced
groupsmeanthattherearenofirmconclusionstobedrawn,butitappearsthat
asinthenormalgroups,participantsthatseeframingdataspendlongerwithit
thanifunframed(U=592,Z=‐1.729,p=.084).
Figure6.29:Meanadjustedtimespentreadingtweets(yaxis)byquestionset(x
axis)forframedandneutralextremists.Errorbarsshowstandarderror.
147
Thenatureofthisdatasetisthat,asnoted,itisincomplete,uncontrolledand
potentiallyunreliable.Certainlyitwillnotbeuseasthebasisforanyin‐depth
analysisforthesereasons.Thereareindicatorsthatthiscouldformthebasisfor
futureresearch.
Discussion
Approach
Asevidentfromtheprevioussection,thisexperimentgeneratedalargeamount
ofparsabledata,withtheassociatedlargenumberofpermutationsthatcanbe
considered.Richnessandmultidimensionalitywassomethingthatwas
intentionallypursuedforthisexperimentinthebeliefthatcross‐modal
assessmentisfundamentallymorereliablethansingle‐measureforthissortof
study.Asthissectionwillhopefullybearout,thisapproachhaslargelybeen
justifiedasmultiplesourcescanbeseentocorroborateeachotherwhen
identifyinglargerabstract‐leveloveralltrends.Wheninvestigatingsomething
relativelyunmappedandnotyetunderstoodsuchasthistopic,thebenefitsof
andreasonsforthisapproachshouldbeself‐evident.
However,thisapproachisnotwithoutitsproblems.Dataoverloadisone
exampleofthis,andthedifficultlyinidentifyingclearresultsinaseaofdata.
Typeoneerrorsareanother;thegreaterthenumberofanalysesperformedthe
greaterthechancethatatleastonetestisafalsepositiverises.Inthisanalysisit
isalmostastatisticallyinevitabilitythatonesuchresultexistswithinthedata
presentedduetothesheernumberoftestsperformed.Thissectionwill
thereforeattempttoaddressthesetwoissuesbynotgenerallyfocusingon
individualresults,andratherbeconcernedwithdrawingouttheoveralltrends
andlargerimplicationsthatcanbeseenfrommultiplemeasures.Thisshould
servetoclarifyanypotentialdataoverload,andtonotplaceunduesignificance
onanyonefinding,avoidingdrawingconclusionsfromasinglefalsepositive.
HypothesisTesting
Beforeaddressingthespecifichypothesisstatedearlier,itisworthaddressing
theimplicithypothesisofthethesisasawhole.Specifically,thisexperiment
148
providesfurtherevidencethatframingeffectsbothexistwithinatask‐based
environmentandhaveuniquepropertiesthatmakethemaninterestingspecific
casetobeconsidered,andthatstrategicadaptationcanbedrivenbycontext.Itis
believedatthispointthattheseprincipleshavebeenestablishedbeyonda
reasonabledoubt.
Thefirsthypothesisstatedthatframingwouldstillbeobservedwiththenew
approach.Thedatafromtheexperimentupheldthisexpectation.Despite
changingthetypeofframeemployed(opinionratherthanstatistical),the
positioningoftheframe(separateratherthanintegrated),andthelevelof
ambiguityandethologicalvalidity(atwitterfeedcontainingdifferingopinions)
framingwasshowntocontinuetobepresentinthesamemanneraspreviously
seen–manifestingasdeviatingfromthe‘neutral’baselinebyabroadly
consistentamountinthedirectionoftheframepresented.Additionally
confidencewasalsoseentovarywiththesefactorsasbefore–againsuggesting
thatthisisabroadercognitivephenomenathanbeingconstrainedtoaparticular
framepresentation.Thereremainbroaderissuestoaddressintheareaof
ecologicalvalidity,anditistruethatthisexperimentusedthesamebasic
paradigmasthepreviousonesothereisdoubtlesslyfurtherworktobedone
movingbeyondthatconstraint.Butthevaryingthemostimportantaspects–the
typeofframingandpositioningofit–haveproducedbroadlythesameresults.
Thesecondhypothesisstatedthatframingwouldcontinuetobeseenonceitwas
removedpartwaythroughtheexperiment.Thiswasalsosupported.The
importanceoftheseresultsrelatedirectlytothepreviouschapter’sresults
whereOver‐Framedparticipantsdidnotshowframinginthewaythatthe
Question‐Framedparticipantsdid,andalsohowinthecardgametaskframing
wasshowntoexistasadirectresultofframingthatoccurredatthebeginningof
thetask.Theseresultssupporttheoverallhypothesisthattheseresultscanbe
explainedbyintegration,oralackthereof,oftheframe.
Theseresultshaveshownthatframingcanhaveapersistenteffectinthissortof
paradigmoncethedirectframehasbeenremoved,sothelackofframinginthe
previousexperimentislesslikelytobeamethodologicalquirk,andmorelikely
149
tobetheframegoingun‐integrated.Becauseitwaspresentedintheinstructions
itwasneveractuallymaderelevanttothetaskinthewaythattheother,
effectiveframeswere(includingtheinstructionsinthecardgame,whichwere
task‐basedasopposedtoscene‐setting).Thissupportstheideathatoneofthe
mostimportantpartsofframingispresentationoftheframeinsuchawaythatit
isnotonlyattendedto,butalsoseeninawaythatisrelevanttothetaskdecision
beingconsidered,andtiesintotheimportanceofthemodel’sintegrationstage.
Additionally,itappearsthattheactualimpactofframingisinitiallynoisyfor
boththisexperimentandthepreviousone,suggestingthatitmaytaketimeand
repetitionforthisintegrationtooccur.Finally,theactualeffectsoftheframe
onceremovedweresignificantlyweakerandlesspronouncedthaninthefirst
elevenquestions.Thiscanbeatleastpartiallyexplainedbytheexperimental
construction,butalsoimpliesthatanyframingwillbeweakerwhennotdirectly
reinforced.Sothisalsosupportsthepriorsuggestionfromthepreviouschapter
thattheintroductionoftheframedidhaveaneffect,butthatitwasmoreto
causeconfusionanduncertaintythansystematicallyaffectthedecisionmade.
Thethirdhypothesispredictedthatconfidenceandothermeasureswouldbe
affectedbyframingasseeninthepreviousexperiment,withframedparticipants
beingmoreconfidentthanunframed.Thiswassupportedbythedata.The
complicatedinteractionswillbeexploredindetaillater,butthegeneral
expectationthatframingwouldaffectthingsotherthanjustwhichchoicewas
madewhenmakingadecisionwasbornout.Confidencewasalsoseentobe
affectedbytheremovaloftheframe,andbywhetheraparticipantwentwithor
againsttheframe.Thishassignificantimplicationsforframingeffectsindynamic
andcomplexenvironmentswheremultiplefactorssuchasthiswillbeatplay.
Thefourthhypothesisstatedthatthemid‐experimentframewouldbeexpected
toframefuturedecisionsfromthatpoint.Thiswasnotsupportedbythedata.
Thisfindingdoes,however,actuallytieintothepreviousobservations.Again,it
canbearguedthattheissuehereisintegrationoftheframe,althoughinthiscase
relativeintegration.Informationwasprovidedandframed,butparticipantsdid
notintegrateitintotheirunderstandingordecisionmakingprocess.Thiswas
150
trueevenintheneutralconditionwhichhaduntilthispointbeenexposedtono
framing,soitisnotsimplyacaseofoneframeoverrulinganother.
Theimplicationisthattheinformationwasnotseenasimportantforbasinga
decisionupon,ormoretothepointthatparticipantsalreadyhadtheinformation
thattheywouldbemakingadecisionaboutitwith.Atthatpointallparticipants
hadbeenexposedtoagreatdealofinformation,framedorotherwise,aboutthe
situation.Thisframesimplywasnotsignificantenoughtoguideadecisionone
wayortheother,somethingthatmustbeattributedtothecontextitwasin,
sincetheactualframingwasmodeledontheAsianfluexamplethatisknownto
beeffectiveinthismanner.Sopriorexperienceisgoingtoberelevantinframe
impact,andsufficientexperiencecanmeanthataframeisineffective.Againthis
speakstotheimportanceofintegration.Thisresultiscomplicatedbythe
apparentretrospectiveframingseenintherethinkquestions,however.Itshould
benotedthatthiswasasingle,relativelysmallresultthatneedstobe
investigatedfurther,althoughitdoessuggestoneintriguingpossibility:itmaybe
possibletoretroactivelyframehowpeoplefeelaboutthedecisionsthatthey
havealreadymade.
Iftherewasasingle,overridingconclusiontobedrawnfromtheseresults,it
wouldbethatframingiscomplicated.Admittedlythisissomethingofatruism–
thesheeramountofresearchandworkputintothefieldistestamenttothefact
thatthisisalreadyappreciated,butitbearsfurtherconsideration.Researchhas
generallybeenconcernedwithidentifyingthefactorsthatcanbeframed,what
differentsortsofframingthereare,andhowthesecanbemanipulated.Noneof
thisissimple,butwhatthisdataillustratesisthatevenwhathaspreviouslybeen
understoodisnottheextentofthecomplexityofthearea.Evenwhen
participantsmaygounframed,whenthereisnosignificantdifferenceinthetype
ofchoicemadeasaresultofexposuretoaframe,participantscanstillbe
affectedbyframes–intheirconfidence,intheamountoftimetheytakewitha
taskand(itmustbepresumed)inotherdimensionsthatthisexperimentdidnot
measure.
151
BroaderThemes
Havingaddressedsomespecificquestions,thisexperimentalsoraisesanumber
oflargerissues.Theresultsofthisexperimentactuallygobeyondspecific
findings;whatisnotableisthefactthatdifferentmeasurescorroborateeach
other.Thesespeaktolarger,broaderthemesthatcanbedrawnfromthework.
Broadlyspeakingthesethemesare:SocialMediaasaframingdevice,Framing
EffectsbeyonddecisionsandFramingEffectPersistence.
SocialMediaasaFramingDevice
Asnotedabove,themainhypothesisofthisexperimentwasnotdisproven:
participantswereframedbytheusedofsocialmediainanongoingtask‐based
environment.Thisframingwasconsistent,didnotdecaywithfeedback,andonly
lessenedwhenremoved.
Superficiallyitmightappearthattheseresultsshownothingmorethanthings
thathavebeendemonstratedwithadvertisingoverthelastsixtyyearsorso:that
people’sdecisionscanbeinfluencedbytherightinformation.Buttodrawthat
conclusionwouldbetomisswhatisspecificaboutframesandparticularly
applicabletosocialmedia.Framesarenotconsciouslyunderstoodtobebiasing;
theyareinternalizedandrationalizedtobetheactor’sownlogicaldecision–
somethingthateventhemostsophisticatedadvertsdonotgenerallyachieve.At
theirbest,peoplearestillconsciouslyawarethatsomeoneistryingtoinfluence
themfromthenatureoftheinteractionwithanadvert.Additionally,thiscontent
andthemannerinwhichitisgeneratedis‘crowdsourced’andorganicrather
thancentralizedanddirected,asisthecasewithadvertising.Framesarenot
simplyopinionsorinformation,butoftenperspectives,interpretationsand
philosophies.Thisresearchillustratesthingsaboutthehowcommunity‐
generatedopinionscanaffectdecisionmakingandspurbehaviours.
Theseresults,particularlytheevidenceofpersistingalteredopinionsoverthe
courseofthetasksuggestthatwecanstarttounderstandsomeofthepreviously
citedreal‐worldexamplesofactivity.Aconventionalmodelofinitiatingsuch
behaviourwouldrelyuponcentralcoordination,buttheaboveresultssuggest
thatduetonewtechnology,userscannowspontaneouslygeneratesuch
152
behaviour.ItisnothardtoseetheLondonriotsinthislight,wheremessages
spreadontheBBMnetworkhavebeensuggestedtoplayalargerolein
organizingthedisruptions.Aquestionthathasrepeatedlybeenaskedis‘whydid
peopleactlikethis’andthisresearchsuggeststhattheanswerlies,atleast
partially,inframingofthedecisions.Theeffectmightbesmall–asthelatter
removed‐framequestionssuggested–butitwouldonlytakeasmallpercentage
ofthethousandsthatsawthemessagestoswellnumbersoverthetippingpoint
beyondwhichthebehaviourbecomesself‐perpetuating.Wealreadyknowfrom
othersourcesofresearchthatpeoplethinkdifferentlyinagrouptotheiractions
asindividuals,andthatthecompulsionto‘gowiththegroup’isareal
psychologicaleffect(Esser,1998).Framedmessagesofthismannercouldbe
easilyseenascreatinga‘group’dynamicfordecisions,andonceenoughpeople
hadmadethechoicetoparticipantthateffectwouldonlysnowballwiththe
visibleevidenceofotherstakingpart.Inmanywaysitissimilartomore
conventionalpoliticalmovements,where‘momentum’isseenaskey.In2008,
BarackObamawasnotthefrontrunnerfortheDemocraticnominationfor
president,butearlyvictoriesinparticularlyIowapushedthenarrativethathe
wasarealcontender,atwhichmoreandmorepeopledecidedtostartdonating
andvolunteeringforhiscampaign(ascanbeseeninthepublicallyreportedFEC
filingsforwhendonationsspiked).Itisnothardtoseefromthishowslight
alterationsinbehaviourchoicesduetosocialmediaframescouldpotentially
haveveryrealconsequences.
Knowingthatsocialmediacanbeframingshouldbeanimportantfactorin
understandingpatternsofinfluenceandeffectwithinthesenetworks.Thiswork
suggeststhatpurelylogicalinformationandconnectionsareinsufficientin
understandingtheseinteractions.Semanticsearchforkeywords,forinstance,
saysnothingaboutthecontextinwhichthosewordsareused.Moreover,
framing–whichasdemonstratedcanaffectplanninganddecisionmakinglet
aloneopinionformation–isallaboutcontext.Aglassbeinghalffull/halfempty
isanobviouslyframedstatement,butsimplycheckingfor‘glass’asatrend
wouldeithermisstheuseinwhichitisbeingapplied,ortheothersideofthe
argument.Similarly,astorycanbewidelyreported,butthemannerinwhichitis
153
reportedtopeoplewillbeofasmuchimportanceastheinformativecontentthat
itpossesses.Indeed,sinceoneofthelimitationsoftwitteristhe140character
limitthatprecludeslongposts,thisbecomesevenmorerelevant.Althoughlinks
tolongerarticlesarewidelyusedonthesite,thereisnoguaranteethatpeople
willactuallyclickonthese.Peoplewhoarefollowingthefeedwillseethetweet
however,andperhapsmostsignificantly,peoplethatdonotcaretoclickona
furtherlinkwouldonlybeexposedtotheframe,whilstalsoimplicitlypickingup
thattheinterpretationisbackedupbywellsourcedmaterial.
Havingobservedthatsocialmediabothcanbeframing,andalsopresents
numerousstructuralandculturalcharacteristicsforenablingthattooccur,a
questionarisesofthepotentiallymoreethicalnature;oneofitbeingusedfor
controlorinfluence.Framingcannotonlyinfluencepeopletomakeaparticular
choices,butwilldosoinwayswherethepeopleframedthenpost‐hoc
rationalizeitasbothrationallybasedandself‐motivated–thepotentialfor
abuseisself‐evident.Onpausingtoconsiderthishowever,itbecomesapparent
thatwhilstthepotentialisthere,therearealsosignificantstructuralbarriersto
itbeingactuallyusedinsuchamanner.
Peopleonsocialmediasiteschoosewhomtheytalktoandwhattheyfollow:
theycancontroltheirviewinghabits.Attemptstoforcepeopletopayattention
tosomethingthattheydonotwanttointhosesituationsoftencomesacrossas
forcedandthusrejectedratherthanwillinglyengagedwith.Sinceoneofthe
prerequisitesofframingisthatparticipantsdonotseetheframeasbeingbiased
orslantedthisisanobviouschallengetobluntattemptsatpersuasion.Twitter’s
useof‘promoted’trendingtopicsforadvertisingrevenueillustratesthis:users
dogenerallyengagewiththetopic,butnotalwaysinaflatteringorbeneficial
light.Heavy‐handedattemptstoframeasituationorinformationwillbeexposed
tothesameissues.
Additionally,thesetupinthisexperimentactuallydemonstratedthatsingular
examplesofaparticularframedonotstopdecisionsbeingpulledintheopposite
directionviatheweightofopinion.The‘counter’tweetframingintheopposite
directionwaspresentinallquestions,butframingstilloccurredregardless.Ifa
154
singularpieceofframingwereattemptedtowardsaparticularend,thedatahere
suggeststhatthiswouldnotbesufficienttoalterbehaviourifotheropinions
werealsopresent.Framedinformationwouldhavetobeplantedonalarger
scalefrommultiplesources,aproblemthatthenoverlapswiththeissueofuser
control.Gettingpeopletofollowenoughsourcesthatwereproducingenough
consistentinformationinordertocreatedsignificantinfluencewouldbe
difficult.Thismaygosomewaytoexplainingsomeofthetroublethat
authoritarianregimeshaveincontrollingsocialmediainthesamewaythatthey
dotraditionaloutlets,ashasbeenseenintheaforementioned‘ArabSpring’.
Becauseofthelackofcentralizedcontrolandprimacyofuser‐directedattention,
understandingtheroleofframinginsocialmediamaybemoreeffectivefor
trackingandforecastingthanforinfluencing.
Contrarily,however,itdoessuggestthatthereisthepotentialforsingle,
influential,andtrusteduserstobepowerfulframedrivers.Populartwitterusers
canamasshundredsofthousands(orevenmillions)offollowers,andtheir
activitycanalreadybeseentoshapeopinion.Sometimesknownasthe
‘twitterati’,anexampleiscomedianandactorStephenFrywhosetweetshave
beenassociatedwiththefailureoftheBlackBerryStormasahandsetinthe
UK(Cellan‐Jones,2008).Onamoreregularlevel,itiscommontoobserve
celebritymentionsengenderingtheadoptionofparticular‘hashtags’orcreating
trendingtopicsasaresultoftheirtweets.Combiningthisresearchwithan
understandingofthesignificantpeopleontwitter(oranothersocialnetworking
site)throughnodetheoryandyoucouldpossiblygeneratetheillusionofan
‘organic’onlinereactionsufficienttoactasaframe.
Theconfidencedatagivessomemoreinsightintohowthisexperiment’sfindings
mayapplyintherealworld.Thisexperimentemployedaforced‐choice
dichotomy,butgenerallyineverydaylifeitiscommonforchoicestohave
multipleoptions,evenifmanyofthemgounusedorunconsidered.Ifnothing
else,mostchoiceshaveatleastthethirdoptionofsimplydoingnothingatall.
Theconfidencedatacangiveusanindicationofhowparticipantsfeltabouttheir
choicesbeyondtheoptiontheychose,implyinghowlikelytheymightbeto
155
actuallydothatthing,whatlevelofefforttheymightputintoitortheirgeneral
attitudeiffacedwiththatproblem.
Firstlyitseemsclearthatcontextiskey.Individualquestionsproducedpolar‐
oppositetrendsinconfidencebothacrossallparticipantsandcomparing
betweenconditions,suggestingthatwhilstthebehaviourthatmaybeprimedby
socialmedia,implementationwillvarywithwhatisbeingasked,whenandwhy.
Thisservesasareminderthatwearestilldiscussingaffecting‘realchoices’–
wherepeoplefeeltheyhavearealdecisiontobemade.Framingandother
factorswillnotbeaspotentifthedecisionisclearlylopsidedortheparticipant
hasstrongpreexistingbeliefs,asthedatafromtheextremistsshowedus.Of
course,importantdecisionsareinvariablythosethataredifficultbecausethere
aremultipleviableoptions,sothisshouldnotbetakenassuggestingthatthis
dataisoflimitedvalue.
Confidencedatagivesussomeideaofhowlikelypeoplearetopursuethe
choicestheymakeinthisexperiment.Forinstance,itseemsthatifapersonis
framedinaparticularway,theywillloseconfidenceintheirdecisionswhenthis
frameisremoved–eveniftheconsensus‘goodchoice’isstilltheonetheywere
previouslypursuing.Morebroadlyhowever,thissuggeststhatframesare
importantevenifpeopleseemunaffected.BetweentheA‐frameandNeutral
conditionsinthelast5(unframed)questionsthereisnosignificantdifferencein
thechoicesmade–butthereisasignificantoneinconfidence.Theframeis
affectingaparticipantbeyondtheactualchoiceobserved‐participantattitudes
havebeenaffected.Thisimpliesthatframesfromsocialnetworkscouldhavean
impactonactionsinwaysthatarenotnecessarilymeasuredbysimply
monitoringchoicesmade,andthattheimportanceofframingmayexistinthe
shiftingoftheseattitudeswhilstnovisibledifferenceisbeingmadetoactual
behaviourinitially.Longbeforeactionsareactuallytaken,onlineframingcould
changemoods,shiftperceptionsandsetthegroundworkforthepointatwhich
actualbehavioursalter–anideareflectedinthevolatilityoftheconfidence.
Whereotherwiseconfidencewasrelativelysmooth,ifframeditbecame
significantlymorevariable.Thebehaviourseenintheexamplesdiscussedreflect
this‐inbothcases,passionsweresuddenlyignitedunexpectedly,andhave
156
generallyleftcommentatorswonderingwhyandhowthiscouldsuddenlyoccur.
Theconfidencedataheredoesnotdirectlyspeaktoallthepotentialcausesof
these,butitdoesindicatethatframesinteractingwithdecisionsraisethe
possibilityoftheseunexpectedlyrapidshiftsinopinionandconfidence,which
wouldseemtobeacharacteristicofthesegrassroots,decentralizedmovements.
Afinalfactorthatshouldbeconsideredwithalltheseobservationsishowthe
datashowsalastingeffect,forbothframeandconfidence,aftertheframehad
beenremoved.Theseeffectsarenotsimplyconditionalonproximateexposure.
Itcannotbeassumedthatsimplybecauseapersonorgroupisnotengagingwith
socialmediaatthattimemeansthattheywillbeunaffectedbyit.Eventheactof
havingbeenremovedcouldbeasignificantfactor–seethedropinconfidence
whentheframewastakenawayintheexperiment.Addtothisthegrowing
modernpervasivenessofsocialmediaingeneral,andthepotentialforsimple
accesstothisinformationtodrive,changeandgeneratebehaviourisclear.
FramingEffectsBeyondDecisions
Asthefinalpointfromtheprevioussectiontouchedupon,oneofthemost
significanteffectstobeobservedinthisstudywasconfirmationofthepreviously
observedeffectthatframingcanhaveonnon‐decisionfactors.Thatisthatwhere
researchintoframingeffectshasnaturallycenteredonthewaythattheycan
swayadecisioninaparticulardirection,thisseriesofstudieshasillustratedthat
thisisnotthelimitoftheimpacttheyhaveonparticipants.
Thiscanbeseeninseveraldatasourcesforthisstudy.Firstly,ashasalready
beennoted,confidencewasaffectedaccordingtobeingframedorunframed.
Effectscouldbeseenbetweenframesaccordingtowhetherchoiceswere
supportedoropposedbytheframeinthefirst11questions,sothereissome
evidencethatthisisaboutdirectionaleffects,butatthesametimeauniform
effectcouldbeseenbetweenframedandneutralparticipantsinthelast5
questions.Inotherwords,framingitself–thesimpleactofframingregardlessof
thedirectionorotherfactors–canimpactaparticipant’sconfidence.
Confidenceisaparticularlyrelevantmetrictoconsiderinthissortof
experimentalsetupandtheoreticalquestion.Asanexampleofwhy,consideran
157
imaginarygorge,spannedbyaricketyropeladderofthesortseeninpulp
adventuressuchas‘IndianaJones’.Inourexample,bothIndyandhiscompanion
forthisadventurechoosetocrossthebridge,buttheydonotsharethesame
amountofconfidenceabouttheirEndeavour.WhereIndystridesconfidently
across,hiscompanionhasdoubtsandproceedsagreatdealslower.Clearly,in
thisexample,bothpeoplehavemadethedecision–tocrossthespanusingthe
ropebridge–buttheirdifferentlevelsofconfidencemeanthattheyapproach
thesameprobleminadifferentmanner.Similarly,ahypotheticalsoldier
operatinginawarzone,maybetoldtocheckaseriesofbuildingsforenemy
combatants.Intelligenceinformshimthatthisareahasnot(unlikemanyareas)
beenseededwithlandmines.Whetherhebelievesthatintelligenceornotwill
clearlyaffecthowhisapproachesthesweepofthebuildings,andthedangershe
prioritizesbeingawareof.Thesamedecisionismade,inbothcases.Butthe
mannerinwhichthatdecisionisimplementedvariesaccordingtoconfidence.
Thisbecomesdoublyimportantwhenwerememberthatvariancecanbeseenin
confidencelevelswhenthereisnodifferenceinthedecisionbeingmade.Bothin
thelastfivequestions,butalsointhequestion‐by‐questiondatafromthefirst11
questionswheredecisionswerestillbeingframed,thereissignificantvariance
inconfidencecomparedtoifthedataisunframed.Sowhetheraparticipantis
framedornotbecomesinsomewaysirrelevant–theircognitiveprocessingwill
havebeenaffectedbytheveryfactthatsomeoftheirdatawasframedandeven
makingthesamechoicestheyarelikelytoresponddifferentlytoanunframed
participant.
Andthiseffectisnotlimitedtoconfidence.Similarly,effectscanbeseeninthe
differencebetweenthetimetakeninreadingthedecisiontext.Here,participants
arereactingdifferentlytoidenticaltext,asaresultofhavingbeenframed,and
participantsalsoreturntoframedinformationmorehavingreaditonce.The
resulthereisthattheactofframingsomethingcanbeseentoaffectthewayin
whichanactiscarriedout:moretimeisdedicatedtothinkingaboutthe
problem,thetimespentondifferentsectionsofthetaskisdifferentand
differentlydistributedasapercentageoftheentiretime.Again,thisprovides
moresupportingevidencetotheargumentthatframeshaveanimpactbeyond
158
thedecisionitself.Iftheonlyplacethattheeffectwasseenwasintheconfidence
measureitmightbepossibletoarguethatitwasanartifactofthatsingle
measure.Evidenceofabroadertendencyhoweverwouldseemtodisprovethat
theory.Evidenceofdifferenttimingpatternsseeninotherwiseidenticaltext
suggeststhatframingisaffectingthewayinwhichthetaskisbeingorganized
andattentionprioritized.Thereseemstobeevidenceofdifferentstructural
approachestothetaskasaresultofframing‐andonethatpersistsafterthe
frameisremoved.
Itispossiblethatpeoplecouldbepresentedwithframedinformationandstill
makethesamechoicethattheywouldhaveinanunframedscenario.Obviously,
wecannotidentifywhatthesechoiceswouldbe–bydefinitionwecannot
observebothaparticipant’sframedandunframedchoices.However,thefactthe
framingdidnotresultina100%choicefavouringoneortheotherdirection,itis
logicaltosuggestthattherearesomedecisionsthatwereunaffectedbythe
framing.Itistobeexpectedthatthiscouldbetrueforareal‐worldtaskaswell.
However,theevidencepresentedheresuggeststhatonsomelevelitdoesnot
matteriftheactualdecisionisaffectedbytheframing.Thesimpleactofhaving
beenframedwillaffectpeople’sperformanceinataskinotherways.Onapurely
cognitivelevel,decisionmakingcantakelongereventhoughthedecisionmade
doesnotchange.
Thisexperimentdemonstrates,essentially,thatframingcanhaveaneffecton
processaswellasoutcome.Andthereareanumberoftasksthatrelyonthis
process,andwillbeaffectedbyitifataskisframed.Anytaskwherespeedand
quickdecisionmakingareimportantforinstance.Indeed,itshouldbe
rememberedthatframingisnotaneither/orpropositioninthesensethatitis
eitherthereornow.Allinformationisframedtoonedegreeortheother;even
‘neutral’informationisdistinguishedbyitslackofframeandactuallyquitea
difficultstatetoachieve.Amajorityoftheinformationweencounterandshare
withotherpeoplewillbeframedbyitsandourverynature.Humanstendto
impartinformationwithqualifiersandsecondaryinformationatthesametime,
informedbyouropinionofagiventopic.
159
Consequentlytheimplicationsofthisexperimentarenotsosimpleastobedealt
withas‐and‐whenframingispresent;mostinteractionswillinvolveframed
informationinsomecapacity.Whatthisevidencesuggestsisthattherearea
multitudeofsituationswhereitcouldbebeneficialtodesignfortheeffectsthat
framingwillbeknowntohaveontasksbeingperformed.Itcouldbeusedtonot
simplyenhancedecisionmakingorleadinaparticulardirection,buttoguidethe
methodologyemployedinimplementationofplansalso.
FramingEffectPersistence
Oneoftherecurringobservationsthatthisthesishasmadeistodowiththe
permanenceofframingbeyondasinglequestion,andthisexperimentprovides
furtherevidenceforthat.Previousresearchhasgenerallyassumedthequestion
offramingrelatesdirectlytowhatisbeingframed–indeed,mostoftheclassic
demonstrationsofframingconcernsingledecisionparadigmsthat,bydefinition,
ceasetocontinueonceadecisionisgiven.
Thisexperimentdemonstratesthatthatviewisinsufficienttoaccountforthe
impactofdecisionframing.Thatmultipleaspectsofcognitiveprocessingbeyond
justthedecisioncanbeaffectedhasalreadybeendiscussed,butadditionally
participantscontinuetobeinfluencedinthedecisionsthattheymakeoncethe
frameisremoved.Confidenceisaffectedasadirectresultofhavingbeen
previouslyframed,decisionstakelonger,sourcesofframeddataareregarded
longerandmorefrequently.Alloftheseresultpoint,indifferentways,tothe
sameconclusion:thatframingeffectsdonotstopatametaphorical‘water’s
edge’ofaquestiontheyrelateto.
Thisisnotsurprisingtosomedegree,butitisneverthelesssignificantbecauseof
theconsistencyandwidthoftheeffectsthatcanbeobserved.Firstly,prior
framingaffectsfuturerelateddecisions.Whilstsomewhatexpected,thisitworth
makingparticularnoteofbecauseitestablishesthattheframeinformationis
beingintegratedintoalargeroverallmentalmodelofaproblem.Ratherthan
beingcontext‐specificandnon‐transferable(asmighthavebeenassumed)it
seemsthatitisbeingestablishedasrelevantgeneralknowledgeandcontext.In
otherwords,theframeisaffectinghowtheinformationitconcernsis
160
understoodandencodedforfutureuse.Giventhatframesareunderstoodto
generallybeundetectedbythepeoplethatarebeingframed,thissuggeststhat
long‐termunderstandingcanbeaffectedbyframing.
Again,thisissomewhataswouldbeexpected.Thenatureofinformationand
learningissuchthatitwouldbesurprisingifpeopledidnotintegrate
understandingbasedonthewaythattheyhadexperiencedit.Regardlessthough,
thisprovidesevidencethatframingisafactorforboththeencodingand
understandingofinformation,andalsothatitmattersinataskbased
environment.Framingsomethingforoneinstancewillaffectfuturedecisions
madeaboutthatsamefactor.
However,somecautionisdueheretoo.Whilstthereweredifferencestosupport
thathypothesis,theyweresmalldifferences.Thegeneraltrendofthedecisions
madewastowardsQuetiaregardlessofhowthequestionswereframed(ornot
framed)uptothatpoint.Framesshouldnotbeseenasbeingsuperordinateto
contextandothertendencies.Indeed,thequestion‐by‐questionanalysisofthe
first11questionchoicesshowsthatframeswerenotsettingthedirectionthata
givenchoicewentin,butratherstretchingthegeneralpatternthatalready
existedinthatdirection.Thissameprinciple–guidingbutnotdefining–canbe
seeninthesmall‐but‐significantdifferencesofthelastfivequestions.
Regardless;thesesmallbutsubtledifferencescontinuetoexist.Asaresultof
framing,futurechoicesareaffected,asisconfidenceandtimespentondifferent
sectionsoftheexperiment.Framesmaybeacaseofsubtleguidanceratherthan
massconversion,butinsomewaysthatmakesthediscoverymoresignificant.
Again,framesaregenerallyinternalizedasbeingaperson’sownchoiceand
opiniononcefollowed,andtheydonotrelyuponalogicalargumenttopersuade.
Thismeansthatpeopleexposedtoframingofaparticulardirectionwillcontinue
tobeinclinedtowardsthatdirection,butmoreduetofeelingsomethingrather
thanthinkingit.Andthismaywellmaketheiropinionshardertomoveinthe
future.Itcanberelativelyeasytologicallydisproveafalseassumption,butmuch
hardertodisputesomethingfelt‘fromthegut’.This,afterall,isthereasonwhy
peoplehavesuperstitionsandfolktheories.Anexampleofsomethingthatis
161
widelybelievedbutactuallyuntrueisthetheoryof‘hothands’inbasketball,
whereaplayerwhosinksathreepointer(alongrangeandrelativelydifficult
shot)isthoughttobemorelikelytodosointhefuture.Statisticalanalysishas
shownthatthiswidelyheldopinionisactuallyfalse–butitisstillwidelyheld.
Thatframingaffectsthelikelihoodtochoosesomethingandtheideathatit‘feels
right’ratherthanalogicalreasonforpickingitactuallyimpliesthatitmaybea
moreusefulformofpersuasionandopinionsetting.Whilstitseffectsmaywell
besmall,whenintegratedtheyarealsopersistentduetopersonalintegration
andopinion,potentiallyatleast.Obviouslythisparticularpropertyofframing
effectsseemslikely,butdefinitelyrequiresfurtherinvestigation.Having
establishedthatframingpersistsasafactoraffectingmanythingsbeyondthe
questionitself,manysuchquestionspresentthemselvesforfutureinvestigation.
162
CHAPTERSEVEN–AdaptationandFramesinTasks:DiscussionAndSpeculation
Thedrivingconcernbehindthisthesisrightfromthestarthasbeenattempting
tobetterunderstandhowpeoplemakedecisions,andspecificallyhowthey
utilisecontextwhilstdoingthat.Aswithmostresearchofthisnature,thesearch
foranswershasopenednewquestionsandexpandedbothunderstandingofthe
answer,andalsotheproblemspaceasitprogressed.Realprogresshasbeen
seeninbothunderstandinghowdecisionmakingisguidedbycontext,and
wherethelimitsofboundedrationalityasameansofexplainingthisbehaviour
exist.Theintentofthischapteristodrawtogetherthevariousrelatedstrandsof
thisunderstandingintoacohesivewhole.
Firstly,therewillbeabriefrecapofthecontext,issuesandmotivationforthis
research,sothattheworkcanbebetterconsideredasawhole.Themodelthat
wasoriginallyproposedinchapterthreewillthenbere‐introducedandits
usefulnessdiscussed.Themainthemesfromtheresearchwillbepresentedas
extensionstotheconceptsofIntegrationandFormulation.
Fromhere,theimplicationsforboundedrationalityandframingwillbe
consideredandatheoryofboundedrationalityintaskspresented.Implications
andapplicationsforthistheoryandtheworkasawholewillthenbeconsidered
beforeconcludingwithsomesuggestionsforpotentialfuturework.
Introduction
Thisthesishasbeenconcernedwithhowpeoplemakedecisions.Primarilyithas
beenfocusedonhowcontextaffectsdecisionandcaninfluencechoice.Forthat
theconceptofboundedrationalitywasusedtomodelsystematicvariationsin
informationthatdidnotvarylogically.Inthisway,thethesishasusedthetwo
relatedpropertiestoexamineeachother.Byplacingboundedrationalityinthe
domainoftaskperformance,itslimitshavebeenevaluatedandnewfindings
made.Conversely,studyinghowdecisionmakingreliesuponcontexthasutilised
163
theframeworkofboundedrationality.Thetworelatedconceptshavebeenused
toexpandandbetterunderstandeachother.
Thisprocesshasinevitablyinvolvedthenarrowingofscopeandcarefully
controlledexperiments.Thisfocusingofscopeshouldnotbetakentolessenthe
breadthofthefindings,however.Theresultsareconsistentwithidentifying
abstractpropertiesthatshouldremaintruethroughvariedinputand
circumstance.Whatthischapterseekstodoistoexplorehowthetworelated
understandingshavebeenexpanded,andtopresentthatunderstandingasa
whole.Itshouldbenoted,however,thatthequestionsbeingaskedwere
inevitablyinterconnected.Sobetterunderstandingofthemodelofadaptation
comeswithsomeunderstandingoftheroleofboundedrationalitywithinthat,
andviceversafortheunderstandingofboundedrationality.Thetwoare
interrelatedconceptsanditwouldbeamistaketotrytoseethemasbeing
entirelyseparate.Afterall,weareanimalsthatarerationalwithinour
boundaries,adaptingtotheconstraintsofatask.
Approach
Inthecourseofthisinvestigation,awealthofdataandfindingsweregathered,
whichprovidessomeproblemsintermsofunderstandingthedataasawhole
anddrawingcohesivefindingsfromit.
Whensuchalargeamountofdataisproduced,thedangerofmakingatypeone
error(falsepositive)isparticularlyhighforatleastoneresultMostofthetests
usedinthisthesishadaconfidencebaselineof.05.Therefore,oneintwenty
testswouldintheorybedowntochance–andthereweremorethantwenty
analysesdone.Thatsaid,manyoftheresultsreportedfarlowerconfidence
thresholds,andcross‐analysiswithinagivenexperimentshouldalreadyhave
theoreticallyloweredthechanceofthaterrorbeingcommitted.Additionally,the
purposeofthissectionistodrawtheexperimentstogetherandidentifythe
generaltrendsandconsistentobservationsthatrunthroughoutit,further
lesseningthechancethattypeoneerrorscouldbeinfluencingtheconclusions
beingdrawnhere.
164
Aconsequenceofthisapproachisthatsomeparticularresultsseenelsewherein
thethesiswillnotbementionedinthissection.Someofthesearepossibly
artifactsoftheparticularcircumstancesinwhichtheyweregenerated.Others
potentiallyrepresentmoregenerallyapplicablefindings,butrequireadditional
researchbeforetheycanbeattributedtotheseunderlyingphenomena.
Regardless,theiromissionshouldnotbeseenasarejectionofthatwork,but
simplythatfurtherworkisneededtoinvestigatethem.
TheModelofAdaptation
Overall,thereweretwomainthemesthatemergedfromthiswork;evidencefor
howinformationwasprocessedinthefirstplace,andthenhowthatinformation
wasusedtocometoadecision.Theseobservationsmatchedupneatlywiththe
IntegrationandFormulationstagesfoundintheoriginalmodeldetailedabove,
althoughitshouldbestressedatthispointthatthisconnectionwasnotedafter
thetrendshadbeenidentified–thiswasnotacaseofconfirmationbiasatwork.
Thefollowingsectionwilldetailtheexpandedunderstandingofthesetwo
processes.
Figure8.1:Ageneralmodelofcognitiveadaptationtoanovelproblem
165
Throughthisthesis,thegeneralpatternofresultshasbeentosupporttheinitial
hypothesisdevelopedinthepreliminarystudyaboutmodellingadaption.The
basicstagesthatweresuggestedhavelargelyhelduptoscrutinyinprovidinga
basicaccountingofperformance.Theresultsthatwereseencouldbeexplained
intermsofthosestages,anditprovidedausefulframeworktoseparatebetween
thedifferentpartsofprocessingwhenreasoninginatask.
Whatthemodeldoesparticularlywellistoprovideaframeworkuponwhichthe
effectsofframingcanbestudied.Ithelpstoseparateoutthecomponentsofthe
process,andtodrawoutthecomplexitiesthatemerge.Whilstthefourinitial
stagesremainuseful(perhapsunsurprisingly,sinceitwasintentionallybroad)
thedetailofknowledgeinthosesectionshasatthesametimeincreasedgreatly.
Itisthisthatwillbeexaminedinthenextsection,particularlyforthesectionsof
formulationandintegration.
Beforethat,however,twopointsofcriticismshouldbenoted.Firstly,therewill
benoattempttomovebeyondthebasicmodelestablishedinchapterthree–
beyondthedeepenedunderstandingofthestagesdetailedbelow.Thisis
intentional.Thethrustoftheresearchinthisthesiswasnotintendedfora
specific,cognitiveprocessinglevelofcognition,andthedatathatdoesexistto
enablethatsortofreasoningisnotcomprehensiveenoughtojustifyaspecific
retooling.Itisbelievedfromthedatathattheaboveiscertainlyanover‐
simplification,butfurtherresearchisneededtoidentifyexactlyhowthe
modulesinteract,ormightbemorecorrectlyconstructed.Thereisalsoawealth
ofadditionalcognitiveresearchthathasalreadybeendone,whichcouldserveto
aidthatprocess,whichfallsatleastalittleoutsidethecurrentscopeofthis
work.Itisenvisionedthatfutureresearchwouldbeabletore‐applythisgeneral
understandingintomorespecificmodellingcriteria.
Secondly,noexpansionwillbemadeoftwoofthestages:feedbackand
implementation.Thatisnottosaythattheyarenotimportant,butsimplythat
thisthesisdidnotprovideenoughinformationtojustifyabroadrethinkofthese
generalcategories.Indeed,muchresearchalreadyexistsforthequestionsthat
couldexpandtheseareas:signaldetectiontheorywithframing,andmotor
166
controltheoriesinimplementationtonametwo.Itwassimplydecidedto
restrictthefocusofthisworktothestagesthattheevidencefromthisthesis
saysthemostabout.Again,itisanticipatedthatfutureresearchandinputfrom
otherdisciplinescouldhelptoexpandthosebetterthanthisthesisusefully
could.
Integration
Thefirstconsistentlyobservablefactortoemergefromthestudieswasthe
importanceandunanticipatedcomplexityofintegration.Integrationasitwas
initiallyenvisionedinthisthesisreferredtotheprocessofdiscriminationthat
occursbetweenperceptionandprocessing,determiningwhatperceivable
informationisthenmadeavailabletobeprocessed.Itwasseenasacognitive
mechanismfortakingtheamountofinformationthatisavailableintheworld
andfilteringitdowntothatwhichisusedfordecisionmaking.Itisapparent
fromcommonexperiencethatpeoplearecapableofmissingthingsthatareright
infrontoftheireyes:tosomehowmissthatpenthey’relookingforevenasit
restsonthedeskinfrontofthem,ortheclassicexampleofsearchingfruitlessly
forapairofglassesthatarealreadybeingworn.Integrationwouldbethestage
atwhichthisoccurswheretheperceivedinformationisencodedandpassedon
(ornot).
Thatthisstagewouldbeimportantsoundslikesimpleconfirmationofthat
whichwouldbeexpectedfromcommonintuition;inordertohaveaneffect
informationmustbydefinitionbeunderstoodinthefirstplace.Howeverwhat
wasanticipatedwasthatthisprocesswasneutral,systematicandobjective,at
leastinregardstosemanticinformation.Sincethesystematicdifferencesinthis
setofstudieswereprimarilyattheconceptualratherthanperceptuallevelit
wasbelievedthatthiswouldhavenoeffectontheprocessofintegration.
Consistentevidence,however,pointstotheideathatintegrationextended
beyondthissimplisticrepresentationandnecessitatesarethinkingofthe
process.
167
ExperimentalEvidence
Thefirstevidencethatintegrationwasimportantformorethansimple
discriminationwasseeninthecardgameexperimentwheredifferentstrategies
wereappliedtoelementsofgameplay.Selectiveapplicationoftheframe
betweendifferentcardtypes(normalandpower)wasobserved,whichwas
interestingonitsownmeritsbutforthepurposesofthissectiondemonstrates
theeffectofintegrationhavingoccurredandthenaframeapplieddifferently
accordingtounderstandingofthetask.Additionally,however,therewasno
effectseeninhowcardswerediscarded.Discardingwasanappropriateplacefor
framingtooccur,andadditionallyalearningeffectoccurredmovingplay
towardsoptimaluse,soitcouldnotbesaidtobesimplybeyondunderstanding
ortooobtuse.Insteaditappearsthattheframinginformationwassimplynot
appliedtothisaspectofthetask.Sotheframehasdefinitelybecomepartofthe
planningprocess,asevidencedbytheeffectselsewhere,butthisisnotageneric
orgeneraleffect,sinceitdoesnotextendtothediscardingrule.Thissuggests
thatintegrationofthediscardingmechanismwasuneven–integratedinto
generalplay,butnotintonovelapplicationsorlearningprocesses.
Thefestivalexperimentthensupportedtheideathatintegrationwasimportant,
albeitinadifferentway.Forthefirstcoupleofquestionstheanswersgivendid
notfallintothepatternthatwaspredictedbyframing,orshowsubstantial
differencesbetweenconditionswhentheydid.Althoughtherewaslater
compliancewiththepredictionsmade,thisinitialperiodofnon‐compliancewith
theframingcanbeexplainedalongsimilarlinesasthecard‐discardingnon‐
adoption:therelevanceoftheinformationthattheframeisprovidingisnot
immediatelyappreciated.Thisgradualadoptioncanactuallybeseentohavea
parallelinthepreviouscardgameexperimentwherethenumberofacesplayed
whenanalternativewasavailableroseinthesecondround,despitethegeneral
trendbeingtowardslessplay.Againthissuggeststhatintegrationoftherelevant
informationwasnotcompleteinitially,andthereforetooksometimetobecome
partofanoverallstrategy.Inbothcasestheinformationoftheframewasnot
immediatelytreatedasrelevanttothedecisionmakingprocess.
168
Thefestivalexperimentprovidedmoreevidencefortheimportanceof
integrationwhencomparingbetweentheOver‐Frameandquestionframe
conditions.Wherethequestionframeparticipantstooktimetointegratethe
frameinformationintoastrategy,theOver‐Frameparticipantsdidnotintegrate
theinformationatall,althoughtheyshowedsignsthattheyknewitwasthere.
Asaresult,theyproducedachaoticpatternforthefirstcoupleofquestions
wheretheirresponseswereunpredictablebeforefallingbackintoaconsistent
patternwiththeneutralcondition–demonstratingthatitwasthisadditional
informationthatmadetheinitialdifference.Again,thisispotentiallyexplained
bytheideathattheinformationwasnotsufficientlyintegratedtoberelevantto
thedecisionmakingprocess.
Finallytheherbalstudyshowedthesamepatternofintegration,takingafew
questionstobeconsistentlyapplied,thatwasseeninthefestivaltask.This
providesconsistencyofobservationbetweenbothexperimentsandframing
types.Additionally,inthefestivalexperimentitalsoshowedlastingstrategic
integrationinthefinalunframedquestions.Thislatterpointisimportant
becauseitsupportswhatwouldbeexpectedbytheconceptofintegrationinto
strategicplanning–thattheeffectwouldstickaroundandberelevantafterthe
specificsupplyingofinformationwasremoved.Particularlybecauseinthis
instanceitwouldbeunlikelytobebiastowardasimplehabit(thatparticipants
hadgottenusedtopickingoneside),astheeventinthatexperimentsuggested
thatabadoutcomehadarisenasaresultofthatveryhabit.Additionally,
integrationhelpstoexplainwhytheeventinthatexperimentdidnothavethe
anticipatedeffectonfuturedecisions–theinformationinitwasnotintegrated–
orattheveryleastnotasmuchasthepreviousframes.Asaresultitwasnon‐
predictiveofresponse.
IntegrationRethought
Whatallthissuggestsisthattoseeintegrationassimplybeingaperceptualand
systematicmatteristooverlookanimportantpartoftheprocess.The
differencesthatarebeingobservedarenotlinkedtoperceptualmechanisms;all
evidencethusfaristhatthatremainsconsistent.Buttheevidencealsosupports
thetheorythatthereisasemanticelementtointegration.Differentstrategic
169
decisionsarebeingmadeaccordingtowhatisunderstoodaboutthetask,and
framescanonlybeappliedwhereitisunderstoodthattheinformationappliesto
thatelementofatask.Integrationisnotsimplytherawinformationthatisthen
passedontoprocessing,butalsothemeansbywhichthatinformationis
processedintoaformthatwereasonwith.Integrationonthislevelisthe
differencebetweenseeingateapotandknowingthat‘thiscanbeusedtomake
tea.’Notethatthisisnotignorance;inallcasestheparticipantsunderstoodthat
theinformationcouldbeappliedtothatarea,itisthatwhendecisionmaking
occurredthatapplicationwasnottakenintoaccount–ithadnotbeen
integrated.
Thisisalsonotsimplyacaseofalearningeffect.Inthecaseofthecardgameit
wasillustratedthatwhathappenswhenaframeisnotinitiallyintegratedand
thenneverre‐presented–ithasnoeffectandnevertookhold,evenwhenthe
discardingmechanicwassubsequentlyunderstood.Intheothertwostudies,
therewasnolearningeffecttobehad.Frameswerenotpredictiveofsuccessnor
didtheypunishnon‐compliance.Ifthisevidencewasduetoalearningeffect
thenthefeedbackshouldactuallyhaveestablishedtheirrelevanceoftheframing
mechanism,butinsteadtheframebecamegraduallymoreinfluentialinaffecting
decisions.Theinformationoftheframewasgraduallybecomingmoreimportant
inthecourseofthetask–itwasbeingintegratedmorecompletely.
Theseresultsdonotprovideafullaccountingofwhyandhowinformationis
integrated,buttheydoestablishbeyondreasonabledoubtthatintegrationisan
importantfactorforframinginongoingtasks.Thequestionthatthisnaturally
raises,therefore,iswhyintegrationhasnotbeennotedasimportanttodatein
single‐choiceparadigms.
Theanswerisactuallyverysimple–thereisalotlessthatcanoccurinasingle
decision.Thereisonequestion,onedecisionandoneresult.Asaconsequence,
thingsareeitherframingortheyarenot,andhavenochancetobeeffective
downthelineorinthefuture.Ifthereisnodiscernableresultastheresultofa
framethenitwouldbeassumedthattheframewassimplynotbiasing–whichis
thereforeoneofthemostimportantthingsthisresearchhastosayforprevious
170
research.Thisexperimenthasestablishedthatframesarenotnecessarily
appliedinstantly,butthatinataskenvironmenttheycanbecomeembedded
afterafewiterations.Theassumptionthathasexistedpreviously‐that
somethingiseitherframingoritisnot‐isdirectlychallenged.Ifthefirstfew
questionsofthefestivalexperimentweretakenontheirown,noresultwouldbe
foundforframinginthequestionframecondition–anullresultusingtheAsian
flumethodologyofframing,arobustandwell‐establishedtechnique.Whatthis
impliesisthattheassumptionthatweknowwhatisframingandwhatisnotmay
verywellbefarfromaccuratebecauseoftherelianceonthesesingle‐decision
mechanisms.Ifframingcanrequiretimetobecomeeffective,ifframingcanwork
overtime,thentheassumptionthatwhatworksinasingledecisionholdstrue
forwhensuchdecisionsaretakeninanongoingdynamicenvironmentseems
self‐evidentlyerroneous.
Anotherissuetheconceptofintegrationraisesisthelackofrangeinmeasuring
effectsinpreviousresearch.Thecardgamestudyshowedthismostpotently
withthecompletelackofframinginthediscardingelementofthetaskand
highlysignificantframingfortheplayingofAces.Whilstinformationwasframed,
andthishadaneffect,itdidnothaveaneffecteverywhere.Integration
anticipatesthissortofresultandprovidesanexplanationforit.Ifthiswerea
single‐decisionexperimentanddiscardingtheonlyvariablebeingmeasured
however,itcouldquiteeasily(anderroneously)havebeenconcludedthatthere
wasnoframingeffectoccurring.Again,seriousquestionshavetoberaisedabout
theappropriatenessofthesingle‐decisionparadigmthathasbeensoprevalent
todate.Notthattheresultsarenecessarilyuntrue(clearly,thebasicprinciples
offraminghavebeenreappliedhere)butratherthatthereisthepotentialthatit
hasbeenguiltyofaseriesoftypetwoerrors,findingnoresultwhereinfact
theremaywellhavebeenonethattheexperimentaldesignswerenotequipped
todetect.
Theenhancedunderstandingofintegrationaddsasignificantdegreeof
complexitytotheconceptofframing.Whilstthesuperficialobservationthatthe
informationneedstobeunderstoodtobeusedisobvious,theideathataframe
canhaveaneffectinoneareabutnotanotherisnot,andtheenhanced
171
understandingoftheprocessesexplainshowthiscanoccur.Framinghas
generallybeenseenasaneither/orproposition,buttheevidencefromthese
studiesisthattherecanbeanelementof‘both’tothatequation.Framingmay
notnecessarilyaffecttheareasthatitisexpectedto,andmayimpactonareas
thatwerenotanticipatedbythesamemeasure.Thekeyhereisthatanullresult
seenfromasinglemetricmaynotindicatealackofframing,oralackofeffect.
Furtherconsideringtheimplicationsofthisforprevioustheory,therearesome
parallelswiththeconceptofaccessibilityasnotedbyKahnemannandTversky.
Accessibilityistheconceptofhowavailablecertainpropertiesareindifferent
contexts;anambiguousfigurethatcaneitherbe‘13’or‘B’giventhesurrounding
contextforinstance(Kahneman,2002).Integrationtouchesuponsomeofthe
sameideas–thatunderstandingcandifferaccordingtohowinformationis
processed–howeveritisdistinctastheyworkatdifferentlevels.Accessibilityis
definedbyinformationthathasbeenintegratedbecausethatinformationis
beingunderstooddifferently.
Whereintegrationdiffersisthatthereisinformationthatcouldbepotentially
biasingbutthatisnotprocessedtoenablethattooccur,orthatanareawhich
couldbebiasedissimilarlynotunderstood.Anyknowledgeorinformationhas
thepotentialtohaveaneffectonaperson’sthinkingorreasoning.Integration
doesnotconcerntheeffectivenessoftheinformationitself,butrathertowhat
degreethatinformationhasbeenunderstoodinordertobeinapositionto
influencethinking.
Consider,forexample,adriverwholivesneartoaschool.Ononesideofthe
schoolistheentrance,andasiscommonthisparticularroadiscoveredwith
slowsignsandwarnings.Thedriverreactstothiscontextualinformationby
slowingtheiraveragespeed,andpayingmoreattentiontocheckforstray
childrenwhomightrunacrosstheroad.Ontheothersideoftheschoolthereare
noslowsignsorwarnings,butthereareholesintheplayingfieldfencesand
hedgeswherechildrenmightseektotakeashortcuteithertoorfromschool.
Wouldthedriverbeexpectedtoslowdownandbecarefulonthisroadtoo?
172
Severalfactorswouldcomeintoplayhere,obviously.Butthebasicbehaviour
canbeexplainedasamatterofintegration.Atarelativelylowlevelof
integration,thedriverwouldnotconsiderslowingdownorpayingmore
attention–theydosointheroadwiththeentranceduetothecontextual
information,butthatinformationhasnotbeenfullyprocessedandasaresult
theyaresimplyreactingtothesurroundingenvironment.Atahigherlevelof
integration,however,adriverwouldpossibilitybeexpectedtotakemorecare
onthesecondroadaswell,becausewhilstthedirectenvironmentisnolonger
cuingthatresponse,theassociationof‘thereisaschoolnearby,Imustslow
down’hasbeenmadeintheirmind.Inthiswaydifferinglevelsofintegrationcan
bepotentiallypredictiveofbehaviouralresponses.
Integrationisthereforeredefinedasbeingthedegreetowhichinformationis
availabletobeusedinformulation,beingconstrainedbybothperceptualand
cognitivelimitations.
Formulation
Thesecondsignificanttrendinthedatathatthisthesispresentsistheeffectof
framingeffectsonfactorsbeyondandindependentofwhatdecisionisreached.
Theimplicationsofthisaresignificant,andprovidetheevidencethatframing
influencesdonotsimplyinfluencetheoutcomeofadecision,butratherthe
mannerinwhichthatdecisionisreachedinthefirstplace‐formulation.Frames
affectdecisionmakingarchitecture,themannerinwhichdecisionsarereached
atall.
ExperimentalEvidence
Thatframeshaveaneffectbeyondsimplyshiftinganopinioncanbeseenacross
thestudies.Primarilyitcanbeobservedintheconfidencemeasuresforboththe
herbalandfestivalexperiments,whereconfidencewasshowninbothcasestobe
directlyaffectedbyframing.Importantlyinbothcasesitwasshowntovaryboth
comparedtoaneutralcondition,andalsowithinframing,whereanotherfactor
interactedwithit–feedbackasthemostnotableinstanceinthefestivalstudy.
Participantsreacteddifferentlytofeedbackasaresultofhavingbeenframed
comparedtotheneutralcondition.Thissuggeststhatitwastheactofframing
173
thatwascreatingand/orenablingthedifferences.Itwasnotthattherewas
simplymoreinformationavailable,sinceparticipantsreactedinadifferent
mannerbyframedespitehavingreceivedthesamelevelofadditional
information.Similarlyinbothexperimentsthepresenceofaframeaffected
decisiontimemeasuresaboveandbeyondanydifferencessimplyassociated
withadditionalwordsoradditionalsemanticcontenttobeprocessed.This
resultisparticularlysignificantbecauseitwasdemonstratedthatthese
systematicdifferencesinadditionalfactorsariseevenwhentheactualdecisions
beingmaderemainedconstant.Notonlythis,butintheherbalstudywherethe
decisionwasseparatedfromtheframingdevice,participantstooklongertoread
thedecisionsectionandthinkaboutthechoicetheyweremaking.Froma
classicalpointofview,framesshouldinfluencetheinclinationofapersonto
makeonechoiceoranother,particularlyopinion‐basedframessuchastheonein
thatexperiment.Noadditional,usefullogicalinformationwasintroducedtothe
processsoitshouldnotbeexpectedtoseeanincreaseinprocessingtimeforthe
decisionitself‐unlessthewayinwhichthatprocessingisoccurringhaschanged.
So,weknowthatframinghasaneffectonfactorsbeyondsimpledecision
making,framingaltersthewaydifferentfactorsinteract,andadditionally
framingcanhaveaneffectonadditionalfactorswithoutactuallyframing
anything.
FormulationRethought
Thisissignificantforseveralreasons.Firstandforemostitissignificantbecause
ittellsusthatthedecisionthatwasmadeisnottheonlythingthatmatterswhen
framingoccurs,andconsequentlyshouldnotbetheonlythingthatismeasured
aswasnotedabove.Anotherimplicationhoweveristhatwhentakentogether
thisdataactuallysuggestssomethingthatgoesfarther‐thatframedinformation
isaffectingthewayinwhichinformationisprocessed.Atraditionalviewofthe
meansbywhichadecisionisreachedisrelativelymechanisticandinput‐output
determined.Arangeofoutputsispossible,butduetodifferentinput,likea
programmablerobotthatwillgotoavarietyofdifferentdestinations.Whatthis
suggestsisthatframingactuallyaltersthemeansbywhichthosesame
instructionsarecarriedout.Inthecaseofthatmetaphoradifferentvehicle
174
altogetherisbeingusedinsteadthatmightstruggletogouparamp,ormore
successfullytraversesomesandasaresult.
Itistherangeofresultsobtainedthatsuggeststhisistrue.Ifinformationwas
beingprocessedinthesamewayinframedandunframedexamplesandsimply
differinginoutput,youwouldexpecttoseeconsistencyintimetaken(after
accountingforwordlengthandsemanticcontent)sincethesamebasic
processingfactorswouldbeperformed.Thatconsistencycanbeseenwithinthe
conditionbetweenquestions,butitcannotbeseenbetweenframedand
unframeddata.Similarlyyouwouldexpecttoseeconsistencyofthereaction
withinconfidencetofeedback.Youmightseeadifferencebetweentheframed
andunframedconditions,buttherelationshipbetweenthedifferentfeedback
typesshouldbeconsistentwhenframedorunframed–unlesstherelevant
informationisbeingtreateddifferentlyasaresultoftheinformationbeing
framed.
Itisforthesereasonsthatitseemsmostlikelytosuggestthatinformationis
actuallybeingprocessedanddealtwithdifferentlyasaresultofbeingframed.
Thatthemannerinwhichtheinformationispresentedhasshapedthemethod
bywhichitisassessedanddisseminated.Rememberthatwithframing,and
indeedboundedrationalityingeneral,thepointisthattheinformationitselfis
generallyofthesamelogicalcontent.Itisthecontext–thebounds–which
resultsinthatinformationbeingassesseddifferently.Whatthisresearch
suggestsisthatitisnotsimplytheconclusionsthatarereachedwhichare
different,butalsothemannerinwhichthoseconclusionsarereachedwhich
varies.
Itshouldbenotedatthispointthatitisunderstoodthattheargumentabout
formulationbeingmadehereisnotconclusivelysupported.Itisbelievedthat
thisevidencemakesthisexplanationthemostlikely,andthatpreviousresearch
supportstheplausibilityofthisframeworkbutadditionalworkisneededto
provideagreaterdegreeofcertainty.However,whatshouldcertainlybeclear
regardlessofthisisthatframeshaveagreatdealofimpactontheformulation
stage.Notjustthatitmighttakeslightlylongertoprocesstheadditional
175
information,butthattherearesubstantialdifferencesinthewayinwhich
informationishandled.Evenifthehypothesisaboutdifferentialprocessingis
notsupported,theimplicationthatframesdohaveaparticulareffecton
processingbeyondthesimpleintroductionofinformationisevidenceofa
significanteffectonitsown.
ReboundedRationality
Thetrendsdetailedaboveshowthatbothintegrationandformulationare
significantlymorecomplicatedascognitiveprocessesthanwasfirstassumed.
However,inusingboundedrationalitytoexploretheconceptofadaptationin
tasks,thatframeworkhasalsobeenusedtomakeobservationsaboutframing
effects,andboundedrationalityasawhole.Whattheevidencesuggestsisthat
thereisaneedtorethinkandreassessboundedrationalityasatheory.
Theoriginalmodel’sinitialpropositionwasthatintegrationdeterminedwhat
wasunderstood,whichthenfedintoplanningprocesseswhichgaveanoutcome
inthemanneroftraditionalinput‐outputmechanisticmodels.Thisresearchhas
demonstratedthatthisisanover‐simplificationofbothprocesses,butmore
significantlyitchallengedseveraloftheassumptionsthattheoriginal
predictionswerebasedon.Framingwasshownnottobeaneither/or
proposition,buttotaketimeinagivencontext.Itwasshownthatsingle
decisionswerenotsufficienttoassesstheimpactofaframeandthatdifferences
occurredoutsidethatmetric.Itwasshownthatfeedbackmatteredandtherefore
thatpreviousexperiencemustalsoberelevant.Alloftheseobservationsrun
contrarytotheimplicitassumptionsofboundedrationalityasawhole,and
framingeffectsspecifically.
Reassessingpreviousworkthemostsignificantflawthatpresentsitselfisthe
overlyreductionistapproach.Amajorityofpriorresearchreliesuponscenarios
constructedwheretwothingsaretrue:thereisasinglechoicetobemadeor
judgmenttoberenderedindiscreteterms,andthatoncethetaskiscompletethe
studyends.Assimplethoughtexperimentsthisapproachisfine,andindeed
therearemanyinstancesineverydaylifewhereadecisioncanberepresentedin
theseterms.Theproblemwiththisapproachis,ofcourse,thatwhilstthat
176
individualdecisioncanwellbemodeledinthatmanner,doingsoremovesthe
complexityofthesurroundingscenariothatalsodeterminestheoutcome.Asan
example,considerthepenaltyshootoutinafootballmatch,andspecificallythe
Englishnationalteam’sperformanceintheminmajorcompetitions.England
havesomethingofareputationforexitingtournamentswhenamatchgoesto
penalties–despitethefactthattheplayersinvolvedintakingtheminevitably
arefamiliarwiththeprocessandfarmoresuccessfulatscoringinpracticeand
wheninactionfortheirclubs.Whydoesthisoccur?Whilstnodefinitiveanswer
canbeoffered,thegeneralassumptionisthatpressurenegativelyaffectsthe
players–especiallysinceonceareputationforlosinginthatmannerhasbeen
establishednervesseemallthemorelikelytokickin.
Thisillustratestheproblemwithareductionistapproachtodecisionmaking.
Theactionsinboththecases(clublevelandinternationals)arethesame,butthe
contextwildlydifferentinawaythatwillproducewildlydifferentresults.And
evenifthesinglepointrepresentationdoestrytocapturethissortofdifference,
thereisstillasignificantgapbetweentakingthefirstandfifthshotsinsucha
contest.Priorinformationisimportant,asiswhatcomesnext.Ultimatelythe
obvioussolutionistoapproachthetaskaspartofawholeratherthanseparate
outitsindividualparts–andthisresearchhasdemonstratedthatbeinginan
ongoingtaskbringsspecificproperties.
Theironyofstatingthisisthatthepointofboundedrationalityhasalwaysbeen
toseedecisionmakingasbeingdeterminedbythesetofcircumstancesitis
madein:cognitive,social,perceptualandsoforth.Whatthisevidencesuggestsis
thatwecanaddanothercategorytothatlist–experimental.Byadoptingthese
standardmethodologiesdecisionmakingstudyitselfhastodatebeen
significantlybounded.
Anoteofcautionshouldbesoundedhere.Boundedrationalityhasshownitself
tobeausefultoolboxforbothprovidingbetterrepresentationsofhuman
decisionmakingthanexistedbeforeit,andalsoforsupplyingthebasisonwhich
decisionmakingwasunderstoodforthiswork.Thisthesisofferssomenew
perspectivesandillustratessignificantareasinwhichthecurrenttheoryis
177
insufficient,butthatshouldnotbemisreadasbeingacompleterepudiationof
theexistingbodyofworkoutofhand.Existingtheoryisausefulbaselinetowork
from–indeed,theframesthatwereusedinthisthesiswereintentionallydrawn
frompreexistingworkforjustthatreason.Asanunderstandingofthemechanics
ofjudgmentsanddecisionmakingitremainsboththeoreticallyandpractically
relevant.Thefactthat,broadlyspeaking,theAsianfluframesuccessfully
influenceddecisionmakinginatask‐basedscenarioactuallyillustratesthatprior
workisnotrenderedirrelevant,simplyshowntobeinsufficient.
RationalityinTasks
Havingestablishedthatitisnecessarytoredefineboundedrationalityforan
expandedworldview,thissectionwillpresentaframeworkfordoingjustthat.
Aswouldbeexpected,thebasiccomponentsofthistheoryhavealreadybeen
explainedintheprevioussectionsandthereforewillnotbeundulyrepeated.
Itisworthnoting,however,thatthisis,andintendstobe,astartingframework.
Thisworkdoesnotpresumetoofferadefinitiveanswertothequestion,butto
furtherunderstandingandenableadditionalknowledgetobepursuedasa
result.Severalareasforexpansionandadditionalworkwillbedetailedinthe
subsequentsectionforpreciselythispurpose.
Thecoreofthistheoreticalre‐structuringisthedivisionofframingeffectsinto
twocomponents;framingandactuation.
Framing
Framingisanyvariationinpresentationofinformationthathasthepotentialto
resultinalternationstotheprocessingofinformationanddecisionmaking.
Previouslyframingreferredtostructuralalterationstoinformationpresentation
thatdidhaveaneffectondecisionmaking.Thisdefinitionexpandstheideaof
framingtoencompass‘non‐effective’framingintentionally.Itremovestheissue
ofsuccessfromadefinitionofframing,andmoreusefullydefinesframingas
simplybeingthevariationwithininformationpresentation.
Actuation
178
Actuationisdefinedasthesuccessorfailureofaframetohaveanimpacton
cognitiveprocessing.Itadoptsmanyofthepropertiespreviouslyassociatedwith
framingasawhole.
Notethatthedefinitionisstilldivorcedfromsimplybeingconcernedwith
decisionmaking,sotheadditionaleffectsandimpactsthatframinghascanstill
beconsideredunderthebannerof‘framing,’astheyshouldbe.
Actuationisitselfdefinedbyanumberoffactors,listedbelow.Thesearethe
primaryfactorsthathavebeenidentifiedbythiswork,althoughitisanticipated
thatthesearenotacomprehensivelistandotherelementsmayalsoemergeas
theresultoffutureresearch.
Potency–Theabilityofaframetoimpactdecisionmaking.
Thisfactorwouldencompassamajorityofpreviousresearch,whichwouldbe
seenasbeingconcernedprimarilywithscenarioswherehighpotencyframes
wereappliedtolimitedtrialenvironments.Italsoenablesthegenerally
appropriateprimaryfocusofmostresearchtobemaintained.Thedistinction
alsoallowsforthepossibilityoflesspowerfulframeswhichmightbeineffective
insingle‐decisionscenarioshavinganeffectwhenimplementedoveralonger
period.
Positioning–Thelocationoftheframerelativetoinformationgathering.Covers
proximitytoframedconcept,aswellasrepetitionandreviewability.
Thisfactorisnecessarytoaccountforthefactthatwhereaframeispresentedis
ofsignificance.Theherbalstudyestablishedthatframesdidnotneedtobe
directlyembeddedintherelevantinformation,buttheworkalsodemonstrated
theimportanceofrepetitionforsomeframes,andthatthesamepositionin
differenttasksmayproducedifferentresults.
Noise–Thelevelofadditionalinformationintheenvironmentthatmayobscure
orotherwisepreventaframefrombeingnoticed.
Thisfactorisreflectiveofthefactthatthecomplexityofthetask,or
environment,canaffectapplicationoftheframe.Thiswasseeninthevariable
179
applicationinthecardgametask.Ittouchesonissuesofattentionand
perceptionaswell,asthesecouldalsopresentbarrierstoframeadoption.
Internalization–Thedegreetowhichaframehasbeenintegratedintoindividual
processing.
Recognizingthatframesretainsomeoftheirinfluenceevenoncetheystopbeing
presented,thisfactormakesadistinctionbetweenaone‐offchoicethatleadto
animplication,andanassociationthathasbeensolidifiedinaparticipant’smind.
EffectScopeThedegreetowhichnon‐primaryeffectsarealsocausedbythe
frame.
Thisfactorconcernstheadditionalmeasuresthatwerenotedbythisthesis.They
aretakenasindicatorsofdifferencesbeingmadeattheprocessinglevel.This
factorisdifferenttotheotherssinceitisobservationalratherthanpredictive,
butimportantastheconfidencemeasuresshowed.Confidenceisnotanticipated
tobetheonlyfactorrelevanttothispoint,butitisusedasanexampleofafactor
thatisbothrelevantanddemonstrablyaffectedbyframing.
Theeffectofthisre‐definitionofframingeffectsisthatthetheorynowprovides
therequireddimensionstoaccountfortheresultsobservedintheworkofthis
thesis.Atthesametimeitdoesnotcontradictoreliminateanypreviouswork.
Allthesingle‐pointdecisionworkthathascomebeforeisstillrelevantandcan
beunderstoodwithinthisframework–however,nowadditionaleffectscanalso
beaccountedfor.
Applications
Thisresearchhasawiderangeofpotentialpracticalapplications.Thereare
threemainwaysinwhichthissectionwilldiscusshowitcouldbeused:
engineeringdesiredbehaviour,avoidingundesirablebehaviourandpredicting
behaviour.Inalloftheseexamplestheprimaryapplicationofthistheoryasit
standsrightnowwouldbeasaseriesofguidelinestobeaddressedbyexpert(or
possiblyinformednovice)specialists.Ratherthanprovidespecificcriteriaof
whatshouldbedoneingivensituations,thetheoryasoutlineaboveprovides
categoriestoassessandaddressasnecessary.Inthiswayitcanbebestthought
180
ofasadiagnostictoolakintocognitivewalkthroughorothersuchusability
guidelines,anditisnoaccidentthattheapplicationofthisresearchissuggested
inbroadguidelinesratherthanspecificremediesorsolutions.
EngineeringBehaviour
Framesandboundedrationalitylosetheireffectivenessiftheyaremadeexplicit
ornoticedbytheparticipantinsomeotherway.Theframesinboththefestival
andherbalstudiesinparticularrelieduponparticipantsacceptingtheunreality
ofthesituationthat,forinstance,the1/3automaticsuccessoftheAsianflu
examplecouldapplytoeverysituation.Thatisnottosuggestthatframesdonot
existintherealworld;theyclearlydoandhavebeenshowntoworkinsuch
situations.Thepointisthattheframesworkedpreciselybecausewithinthe
context(theboundsofthesituation)theyappearedintheywereplausibleand
appropriate.Thechallengeofapplyingthisworktotherealworldisnotonlyin
findingappropriatesituationstouseitin,butalsoinfindingawaytoapplyit
thatdoesnotbecomeself‐defeating.Somethingthatthisthesishasaptly
demonstratedisthattheeffectofframesveryagreatdealaccordingtothe
contextthattheyareplacedin–tasksoversingledecisions,positionandsoforth.
Itiscertainlybeyondthescopeoftheevidenceinthisworktoprovidealistof
exactlyhowandwhyframeapplicationmayvaryinagivencontext,sotheroleof
anexpertbecomesnecessarytotranslatethegeneralabstractunderstanding
intocontext‐specificapplications.
Anexampleofwhereitcouldbeusedtoengineerdesiredbehaviouristheworld
ofonlineadvertisingandselling.Heretheobjectiveisobviouslytohavemore
consumersclickthroughtoawebpagefromanadvert,andthentomake
purchasesoncetheyhavedonethis.Whatmakesthistheoryspecifically
applicabletothisareaisthatonlineinteractiontakesplaceinahighly
engineeredenvironment.Whilstagreatdealoftheinteractioncannotbe
predictedaheadoftime–realworlddistractions,differingmonitorsetups,etc.,
etc.–forthemostparttheenvironmentthatanytwopeopleexperiencewhen
visitingagivenwebpagewillbelargelyidentical.Evenwithpersonalizationmost
sitesruna‘theme’andagivenlayoutthatallusersexperienceinthesamebasic
way.Thisallowsformorepreciseapplicationoftheprinciplesanddesignto
181
emphasizethem.Ofparticularinterestwouldbetheconceptofinternalization,
becauseitissuchadesirableoutcomeiftheobjectiveistosellsomething(and
particularlyforengenderingrepeatcustomers).Tryingtoensurethatthebasic
conceptsofpurchasingtheproductorusingaserviceareinstilledwould
arguablybemoreimportantthanactualclickthroughorsales,initiallyatleast.
Ifitmanagestoimpartahabitorbehaviour,thatwillbefarmoreusefuloverthe
longrunthanaone‐timesuccessfultransaction.
PreventingBehaviour
Acounter‐exampleofwherethissameinformationcouldbeusedtoprevent
mistakesbeingmadeasaresultofframedinformationwouldbehigh‐density
decision‐intensiveenvironmentswithhighconsequences,suchasbattlespace
managementorairtrafficcontrolrooms.Inthesesetupshigh‐skillindividuals
makeconstantdecisionswhichoftencarryhighpenaltiesifamistakeismadeor
afailureoccurs.Ideally,adecisionmakerinthiscontextwouldbeprovidedwith
therelevantinformationinthemostusefulmannerandthenallowedtomake
thatdecisionotherwiseunbiased.Thedangeristhattheinherentbiasesthat
peoplecarrywiththemcouldcausethedecisionstobebiased,ascould
inadvertentframingoftheinformation.Itisnotimplausiblethatinformation
comingintoabattlespacecoulditselfbeframedevenifthatframingis
unintentional.Soldiersconveyinginformationfromafrontlinemaynotbe
reportingwithcompleteobjectivity,afterall.
Inthiscontexttheresearchcouldbeusedasaguideforwhattoavoidinorder
tolessenthechancethatinformationisbeingframedorbiased.Anadvantageof
re‐conceptualizingtheconceptofframinginthewaythathasbeendoneisthatit
treatsallinformationpresentationaspotentiallyleading,andthereforedoesnot
requirethat‘frames’arespecificallyidentifiedtobedealtwith.Ratherthe
factorsinvolvedinactuationcanbespecificallyaddressed.Ofparticularinterest
forthiswouldbeinternalization.Studyingsystemstobeawareofwhether
participantsareinternalizingbiasedchoicesfortheirdecisionmakingisboth
quantifiablyachievable(pastdecisionscanbetrackedagainstprojected
baselines)andshowntobenewlyrelevantbythisresearch.Previoustheory
wouldhaveassumedthatagivendecisionhadthesamechanceofaffectinga
182
choiceinitsfirstor60thiteration,butasthisthesishasshownevenwell
establishedframescantaketimeto‘comeonline.’Thusthetheorywould
supportlookingforongoingandemergingbiases,aswellasdetectingmore
subtleframeeffects,andprovidingsomeinsightintohowtoaddressthese–
simplychangingthewaytheinformationispresentedwillnotbeenoughasby
thatpointthetendencywillbeintegratedintoawayofthinking.
Additionally,thistheorywouldpotentiallybeabletolinkotherphenomenaback
toframing.Aswasdemonstratedinthethesis,framingcanhaveaneffecton
confidenceevenwhentheactualdecisionsbeingtakendonotchange.
Battlespacemanagementrequiresconfidencetoundertake–beinganarenathat
isbydefinitionfilledwithdangerandpotentialflaws,confidenceisanecessary
qualitytogiveappropriateorders,anddangerousbothinitsabsenceandalso
whenthereistoomuch.Overconfidenceorunderconfidencewouldbe
somethingthatcouldbothbenoticed,andthenpotentiallyascribedtothe
framingthatwaspresent.Thissortofsecondaryreaction,andbeingabletolook
foritcouldbepotentiallyimportantsinceitmightrepresentawayto
systematicallyaddresswhatwouldotherwisegenerallybethoughtofasa
‘character’flawratherthananenvironmentaleffect.
PredictingBehaviour
Thethirdmannerinwhichthisresearchcouldbeappliedisinunderstanding,
analyzingandpredictingbehaviourpatterns.Ashasbeenraisedalready
elsewhereinthisthesis,anareawithagreatdealofpotentialforthisiscyber
influence,andspecificallysocialmediaasaconduitforthat.
Withtheinfluencethatsocialmediacurrentlyhasinsociety,andtheexpectation
thattheunderlyingprinciplesthatdriveitwillcontinuetobeinfluentialinyears
tocome,thisisanareawherethereisagreatdealavailabletobeunderstood
andthereisgoodreasontodoso.Whilstthedataitprovidesissomewhathard
toreadattimesduetothesheervolumeanddifficultyassociatedwithpicking
outtherelevantinformation,theactuationfactorslistedaboveprovideagaina
frameworkforbothunderstandingbehaviourandpredictingit.
183
Thisactuallyprovidesagoodexampleofanareawhereadaptabilityiskeyfor
thetheory.Clearlythereareagreatdealofspecificcontextualissuesthatarisein
theareaofcyberinfluence–interactivityofframes(opinion),asynchronous
communication,dynamicenvironment,personaltiesandaffiliationetc.Context
specificityissomethingthatshouldbeembraced,andtheoverarchingprinciples
usedasaframeworkratherthanadefinitiveanswer.Indeed,specificsocial
networksaregoingtohavespecificnormsandinteractionparadigmsthat
conflictwithothersevenutilizingthesameplatformforcommunication.On
twitterthehashtag#tcot(standingfor‘topconservativesontwitter’)is
commonlyusedbyconservativepoliticaloperativesasadefaultsothatother
userscansimplysearchforthattagtoseewhatisdrivingthedayintermsof
storiesonthatside.Thereisnoequivalentexampleinsuchwidespreaduseon
theliberalside,whichtendstoappropriatehashtagsforagivendayandevent.
Thepointisthatthisisaclearorganizationaldifferencebetweentworelatively
similar(politicallydriven)groupsonthesamecommunicationplatform.In
termsofthetheory,theuseofthe#tcothashtagwouldenableaframingtweetto
gainrelativelystrongpositioninginawaythatwouldnotbepossibleina
differentgroup.Contextiskey,butagaininthisexamplethetheoryprovidesa
wayofunderstandingthelikelihoodofthatframingspreadingorbeing
internalized.
Anotherpotentialapplicationofthisworkisthatitcouldbeusedtoactually
identifywhenpeoplehavebeeninfluenced.Oneofthemainissuesthatframing
andtheoriesofcyberinfluenceandotherworkfaceisdeterminingwhenan
attemptedinfluenceisactuallyworking.Forlaboratoryworksuchasthis,
controlgroupsandbaselinescanbeestablished,butinmoreecologicallyvalid
situationsitisoftenimpossibletocomparebetweengroupsinthisway,soitis
hardtoknowifanapplicationisactuallyworkingornot.
Butthisresearchdoesprovideameansoflookingatthatproblem.Ithasshown
thatdecisionmakingisnottheonlyfactorthatcanbeaffectedbyframing,but
alsoconfidenceandreactiontimes.Additionaleffectsarenotablewhen
participantsaresuccessfullyframed,anditistheseadditionaleffectsthatcan
providethebasisforestablishingifaframehasbeensuccessfulornot.Inatask‐
184
basedenvironmentwithhighmonitoring,reactionanddecisiontimescanbe
monitoredforchangesagainstabaselinereadingfrompreviousperformanceon
thattask.Alternatelywhenconsideringanareasuchascyberinfluence,thereis
thepotentialtousethefactthatfeedbackisavailableinthatarea.Peoplearenot
simplyconsuming,butalsobroadcasting,andthesemanticcontentoftheir
publishedcontent,aswellastheirnoticeableactions,couldprovideinsightinto
howtheirconfidencelevelshave(orhavenotbeen)affectedbyexposuretothe
framedmaterial.Inanycase,theexistenceofsecondaryeffectsenables
additionalstreamsofinformationtobeusedinunderstandingifinfluencehas
occurredornot.Theseshouldbeabletobeemployed,alongwithothermethods,
inanattempttotriangulateananswerfrommultidimensionalsources.
Autonomy
Finally,itisworthconsideringtheimplicationsthistheorycouldhaveforthe
designofautonomoussystems.Whilstthereissomeobvioustheoreticaldistance
betweenthisworkandartificialintelligenceatthispoint,atthesametimeitis
believedthatthisframeworkcouldprovidesomebasisforinnovativeworkthere
aswell.
Previouslyframeswereconsideredprimarilyintermsoftheirpotencyas
definedbytheirinteractionwithpeople.Accountingforthisinanartificial
settingwasalwaysgoingtobedifficultbecauseofthecircularnatureofsucha
process–ifyoucanonlydefineaframebyinteractionwithaperson,thena
humanhastobeintheloopforanautonomoussystemtoseeitremovingthe
autonomyfromthecalculation.
Thisworkdoesnotentirelysolvethisissue.Potencyisstillaqualitythatremains
somewhatdefinedbypeople.Howeverotherfactorsofactuationaremore
quantifiable.Thestructureofdifferentialinformationrepresentation,asaframe
isnowdefined,canbeobjectivelymappedanddiscriminatedbetweenonabasic
level.Similarly,positioningcanbedefinedintermsofinformationlocationand
theunitsbetweenthosepoints(pageviews,distanceonascreen)ascansome
environmentalnoise(additionalwindows,colourvariety,etc.)Thiscannot
provideforallthecomplexitythatgoesintoframinganymorethanthetheory
185
pretendstodefinitivelyrepresentthearea.Butwhattheydoprovidearethe
workingbasicsforunitsofframingtobebuiltupandunderstood.Automated
systemsmightnotbeabletoactuallydetectorunderstandframing,butby
developingandapplyingintelligentalgorithmsthatlookforthesequalities–and
specificallycombinationsofthesequalities–thenthereemergesthepossibility
offramingevents,informationorscenariostobedeveloped.Anautonomous
systemcouldmonitorinformationflowandhighlightforahumanoperator
whereinformationmightbeframinganopinionsothatthatoperatorcanmakea
distinctionforthemselves.Framestendtolosepotencyoncetheyarespecifically
lookedfororconsidered,includingsuchexamplesastheAsianfluclassic.By
usingtheseprinciplestomanageinformationflowtoahumanoperator,thereis
thepotentialtoavoidbiasingsaidinformationinawaythatotherwisewouldgo
unnoticed.
FurtherWork
Thisworkredefinesapreviouslywell‐mappedareaoftheory.Consequently,
thereisagreatdealoffollow‐upworkthatcanbeundertakenasaresultofthe
newquestionsitraises.
Itseemsprudenttofirstacknowledgethatthereisagreatdealofbasicfollow‐up
worktobedone.Researchshouldbechallengedandthereshouldbeattemptsto
disproveit,includingtheconclusionsofthisthesis.Subsequentworkinvolving
taskactivity,confidence,andfeedbacktofurtherexaminetherelationships
thereinshouldbewelcomed,althoughthissectionwillnotaddressthesedirectly
outsidethisparagraphinfavourofbroaderideas.Thatsaid,thetheoryprovides
obvioustestabilityfortheseparatefactorsofactuationtobeinvestigated,aswell
asthemoregeneraltheoriesofintegrationandformulation.
Themostobviouslyspeculativetheorydrawninthissectionisthatofframing
causingdifferentcognitivestrategiestobeemployedindecisionmaking.There
hasalreadybeensomeworklookingatbrainactivationpatternsinsingle‐
decisionworkthatsupportsthishypothesisasplausible(Deppeaetal.,2005;
Gonzaleza,Danaa,Koshinob,&Just,2005;Martino,Kumaran,Seymour,&Dolan,
2006)butthereisclearlyneedforfurtherexaminationoftheidea.Therearea
186
numberofpotentialproblemstobeovercomehowever,themostprominentof
whichbeingthatbrainimagingtechniquesgivegenerallybadtemporal
resolutionwithcurrenttechnology–andthisresearchdealswithconsecutive,
temporallycloseoccurrences.Thehypothesisalsoallowsfortheideathat
differentcognitivestrategiescouldemergeinthesameregionofthebrain,sothe
differencebeingsoughtcouldpotentiallybeundetectablewithcurrent
generationmachines,particularfMRI.
Nevertheless,itremainsaninterestingareaofpotentialresearch.Comparisons
couldbemadebetweenbrainsthathavebeenprimedandunprimedbya
framingeffecttolookforadifference,enablingasnapshottopotentiallyshow
deepercognitivedifferences.AlternatelyfMRIcouldbeusedtolocatethe
associatedbrainregions,andthenEEGorequivalentusedfortemporal
examination.Whatisclearinanycircumstancesisthataseriesofstudieswould
benecessarytoaddressthehypothesiswithanycertainty.
Anotherareathatwouldexpandthisresearchinasignificantwaywouldbeto
considerthedimensionofexpertise,andpre‐existingknowledgeandskill.Oneof
thecriteriaappliedthroughoutthisthesiswasthatthetasksshouldbenovelin
someregard;thatparticipantswerelearningastheywentalong,atleast
partially,becausethatisthetimewhenapersonwouldbeexpectedtobemost
opentoinfluencefromframesandothermeansofdirectingopinion.Thisshould
notbeseenasaflaw,sinceitrepresentedthemostappropriatewaytoensurea
levelplayingfieldbetweenparticipants.Expertiseitselfwascarefullycontrolled
forandnosignificanteffectswerefoundassociatedwithit.Butwhilsttheissue
ofexpertiseliesintentionallyoutsidethescopeofthisthesis’work,it
neverthelessrepresentsaninterestingsourceofpotentialwork.Theextremists
thatwereidentifiedintheherbalstudyareanexampleofhowpre‐existing
biasescanoverruleframingeffects;itstandstoreasonthatpre‐existing
experiencethatwaslogicallybasedcouldhaveasimilarimpact.
Thisresearchhasdemonstratedthatfeedbackcanovercomeaframingeffect,
althoughitdoesn’tnecessarilydoso.Experiencewouldbepredicted,therefore,
tofollowasimilarpattern.Ononelevelitwouldbeanticipatedthattask
187
familiaritywouldsimplyoverruleanypotentialframingeffect,sinceexperience
wouldbeprovidingthemethodologybywhichadecisionwasbeingmadeand
presumablybegivenmorecognitiveweightthancurrentinformation.Onthe
other,peoplestilldemonstrablyadapttochangingcircumstancesandrespondto
newinformation,whichwouldpresumablyincludenewinformationthat
happenedtobeframed.Sosomesortofrelationshipshouldbepresentbetween
thetwoconflictingfactors.Experiencewould,presumably,representanother
factorthatwouldinfluenceframeactuation.
Athirdsuggestedareafortheoreticalresearchwouldbetolookatthe
interactionthatappearstooccurbetweenframingandfeedback.Thisfinding
wasshowninthefestivalexperiment,butintentionallynotinvestigatedany
furtherbecausethequestionsitraisesaresignificantenoughthatitcould
potentiallybeabodyofworkitself.Forthepurposesofthisthesisitwas
sufficienttoknowthattherewasarelationshipbetweenthetwofactors,andas
supportingevidencefortheideathatframingaffectedmorethanjustthe
decisionbeingmade.
Theresultraisesgreaterquestions,however.Whywereframedparticipants
moreconfidentwithqualitativefeedback?Thesuggestedhypothesisisthatit
allowsthemtoseetheinformationsubjectively,andthisissomethingthatcould
standtobemoreempiricallytested.Thathypothesisalsomightflowbackinto
theexpectationofdifferencesinprocessing–differenttypesoffeedbackcould
generatedifferentlevelsofconfidence,whichcouldberelatedtothewayin
whichaproblemisconsidered.
Morethanjustthishowever,theresultraisesquestionsaboutfeedback
interactingwithframingingeneral.QualitativeandQuantitativefeedbackare
usefuldistinctions,butarenottheonlywayinwhichfeedbackcanbepresented,
nornecessarilythemostappropriatewayofdistinguishinginformationtype.
Quantityofinformationwouldbeonefactor,butalsoquality.Theinformationin
thelattertwostudiesofthisthesiswasintentionallynon‐informativetoavoid
pullingparticipantstooneparticularconclusion,andtomaintainalevelof
ambiguity.Butwhataboutsituationswherefeedbackismoreexplicitanddirect?
188
TheCardGamestudysuggestedthatthiswouldeventuallyover‐runaframing
effect,althoughthatwasaframelocatedintheinstructions.Wouldarepeated
framemaintainsomeapplicabilityinthefaceofconflictingempiricalfeedback?
Clearlythereareagreatdealofunansweredquestions,andthisworkhas
hopefullyopenedupapathwaytostartaddressingthem.
AFinalWord
Theconclusionsfromthisworkshouldbereadilyapparentatthispoint.Context
andinformationpresentationcaninfluencestrategicformation,andframing
effectshavedistinctpropertieswhenplacedinanongoingtaskparadigm.The
evidencefortheseobservationsseemsindisputableatthispoint,whilstthe
implications,applicationsandfurtherworkdiscussedinthissectionhopefully
provideabasisfortakingthemonwards.
Therehopefullyremainsroomforabriefphilosophicalnotefromtheauthor.
Thisworkwasatleastpartiallyinspiredbyanddrivenfromaninterestin
artificialintelligenceandautonomoussystems.Thereisclearlysomedistance
betweenthatareaandwherethisthesisandconclusionsitcontainshaveended
up,butthesignificanceofthelargerabstractquestionsaboutartificial
intelligenceandautonomyisthattheyarethesamequestionsthatexistabout
humanautonomyandintelligence–howdowethink?Howdowereason?Why
dowechoosetodocertainthingsandnotothers?Thesearenoteasyortrivial
questions.Theyarenotsimpleorprescriptive;theyarenotasimpletasktobe
replicatedorasingleskilltobeaped.Theyarenotemergentpropertiesofa
sufficientnumberofsimplerules.Fewthingssoaptlyillustratethecomplexityof
thehumanmindasthepersistentinabilitytocreateanythingapproachingan
artificialcorrelateofit.Asaconsequencethesamequestionsthathavealways
drivenpsychologystarttobecomethosethatdriveAI,andviceversa.
Sowhilstthebulkofthisthesisdoesnotultimatelyrelatetoautonomyand
artificialintelligenceitishopedthatitmightatleastpartiallyserveasa
reminderoftheimportanceofinterdisciplinaryworkandcollaborationacross
interestareas.Itisnoexaggerationtosaythatthisthesiswouldnotexistwithout
thequestionsthatwereraisedfromexposuretoresearcherswhoseperspective
189
andinterestsweregreatlydivergentfromthosetheauthorstartedwith.Asa
small,anecdotalandquantitivelyunsupportedobservation‐thebenefitof
makinganefforttolistenandthinkoutsideyourcomfortzoneissomethingthat
shouldnotbetakenforgrantedandcanreapunexpectedbenefits.Hopefully,this
thesiscanprovideasmallillustrationofthatpotential.
190
References
Allis,V.(1988).AKnowledgebasedApproachtoConnectFour.Masters,VrijeUniversiteit,Amsterdam.
Anderson,J.R.,Matessa,M.,&Lebiere,C.(1997).ACT‐R:ATheoryofHigherLevelCognitionandItsRelationtoVisualAttentionHumanComputerInteraction12,439‐462.
Asimov,I.(1950).I,Robot.NewYork:GnomePress.Bakshy,E.,Hofman,J.M.,Mason,W.A.,&Watts,D.J.(2011).Everyone’san
Influencer:QuantifyingInfluenceonTwitterPaperpresentedattheWSDM,HongKong,China.
Berliner,H.,&Ebeling,C.(1989).PatternKnowledgeandSearch:TheSUPREMArchitecture.ArtificialIntelligence,38(2),161‐198.
Bertrand,M.,&Mullainathan,S.(2004).AreEmilyandGregMoreEmployableThanLakishaandJamal?AFieldExperimentonLaborMarketDiscriminationAmericanEconomicReview,94(4),991‐1013.
Bibby,P.A.,&Payne,S.J.(1993).InternalisationandtheUseofSpecificityofDeviceKnowledge.HumanComputerInteraction,8,25‐56.
Bowden,M.,Jung‐Beeman,M.,Fleck,J.,&Kounios,J.(2005).Newapproachestode‐mystifyinginsight.TrendsinCognitiveSciences,9,322‐328.
Carraher,T.N.,Schliemann,A.D.,&Carraher,D.W.(1985).MathematicsintheStreetsandintheSchools.BritishJournalofDevelopmentalPsychology,3,21‐29.
Castillo,C.,Mendoza,M.,&Poblete,B.(2011).InformationCredibilityonTwitter.PaperpresentedattheWWW,Hyderabad,India.
Cellan‐Jones,R.(2008).CanStephenFryKillaGadget?Retrieved25/04/2012,fromhttp://www.bbc.co.uk/blogs/technology/2008/11/can_stephen_fry_kill_a_gadget.html
Chenga,F.‐F.,&Wu,C.‐S.(2010).Debiasingtheframingeffect:Theeffectofwarningandinvolvement.DecisionSupportSystems,49(3),328‐334.
Chomsky,N.(1959).AReviewofB.F.Skinner'sVerbalBehavior.Language,35(1),26‐58.
Deppea,M.,Schwindtb,W.,Krämera,J.,Kugelb,H.,Plassmannc,H.,Kenningc,P.,&Ringelsteina,E.B.(2005).Evidenceforaneuralcorrelateofaframingeffect:Bias‐specificactivityintheventromedialprefrontalcortexduringcredibilityjudgments.BrainResearchBulletin,67(5),413‐421.
Dou,W.(2010).Comparingdifferentlevelsofinteractionconstraintsforderivingvisualproblemisomorphs.Paperpresentedatthe2010IEEESymposiumonVisualAnalyticsScienceandTechnology(VAST).
Dreyfus,H.L.(1972).WhatComputersCan'tDo.London:TheMITPress.Esser,J.K.(1998).AliveandWellafter25Years:AReviewofGroupthink
Research.OrganizationalBehaviorandHumanDecisionProcesses,73(2‐3),116‐141.
Evans,J.(2005).DeductiveReasoning.InK.J.M.Holyoak,R.G.(Ed.),TheCambridgeHandbookofThinkingandReasoning.Cambridge:CambridgeUniversityPress.
191
Feather,N.T.(1968).ChangeinConfidenceFollowingSuccessorFailureasaPredictorofSubsequentPerformance.JournalofPersonalityandSocialPsychology,9(1),38‐46.
Francis,L.J.,Lewis,C.A.,&Ziebertz,H.(2006).Theshort‐formrevisedEysenckPersonalityQuestionnaire(EPQ‐S):AGermanedition.SocialBehaviorandPersonality,34(2),197‐204.
Frederick,S.(2005).CognitiveReflectionandDecisionMaking.JournalofEconomicPerspectives,19(4),25‐42.
Gächter,S.,Orzen,H.,Renner,E.,&Starmer,C.(2009).Areexperimentaleconomistspronetoframingeffects?Anaturalfieldexperiment.JournalofEconomicBehavior&Organization,70(3),443‐466.
Gigerenzer,G.(2000).AdaptiveThinking.Oxford:OxfordUniversityPress.Gigerenzer,G.(2008).RationalityforMorals.Oxford:OxfordUniversityPress.Gigerenzer,G.,&Selten,R.(2000).BoundedRationality.Cambridge,MA:TheMIT
Press.Gilbert,D.T.(2002).InferentialCorrection.InT.Gilovich,D.Griffin&D.
Kahneman(Eds.),HeuristicsandBiases.NewYork:CambridgeUniversityPress.
Gobet,F.(1997).Apatternrecognitiontheoryofsearchinexpertproblemsolving.ThinkingandReasoning,3,291‐313.
Gonzaleza,C.,Danaa,J.,Koshinob,H.,&Just,M.(2005).Theframingeffectandriskydecisions:ExaminingcognitivefunctionswithfMRI.JournalofEconomicPsychology,26(1),1‐20.
Halliday,J.(2011).Londonriots:howBlackBerryMessengerplayedakeyrole.Retrieved20thSeptember2011,fromhttp://www.guardian.co.uk/media/2011/aug/08/london‐riots‐facebook‐twitter‐blackberry
Hirschfeld,L.A.,&Gelman,S.A.(1994).MappingtheMind:DomainSpecificityinCognitionandCulture.NewYork:CambridgeUniversityPress.
Hogarth,R.(1987).JudgementandChoice(2nded).Chichester:JohWiley&Sons.Howard,P.(2011).DigitalmediaandtheArabspring.Retrieved20thSeptember
2011,2011,fromhttp://blogs.reuters.com/great‐debate/2011/02/16/digital‐media‐and‐the‐arab‐spring/
Howes,A.,Lewis,R.L.,&Vera,A.(2009).RationalAdaptationUnderTaskandProcessingConstraints:ImplicationsforTestingTheoriesofCognitionandAction.PsychologicalReview,116(4),717‐751.
Huang,Y.,&Wang,L.(2010).Sexdifferencesinframingeffectsacrosstaskdomain.PersonalityandIndividualDifferences,48(5),649‐653.
Kahneman,D.(2002).MapsofBoundedRationality:APerspectiveonIntuitiveJudgementandChoice.LesPrixNobel2002.
Kahneman,D.(2011).Thinking,FastandSlow.London:Penguin.Kahneman,D.,&Tversky,A.(1973).OnthePsychologyofPrediction.
PsychologicalReview,80,237‐251.Kelman,H.C.,&Barclay,J.(1963).TheFScaleasameasureofbreadthof
perspective.TheJournalofAbnormalandSocialPsychology,67(6),608‐615.
Kendall,L.,Hartzler,A.,Klasnja,P.V.,&Pratt,W.(2011).DescriptiveAnalysisofPhysicalActivityConversationsonTwitter.PaperpresentedattheCHI,Vancouver,BC,Canada.
192
Kepner,C.H.,&Tregoe,B.B.(1965).TheRationalManager:ASystematicApproachtoProblemSolvingandDecisionMaking.NewYork:McGraw‐HillBookCompany.
Kim,K.,Proctor,R.W.,&Salvendy,G.(2012).Therelationbetweenusabilityandproductsuccessincellphones.Behaviour&InformationTechnology,31(10),969‐982.
Kirsh,D.(2009).ProblemSolvingandSituatedCognition.InP.Robbins&M.Aydede(Eds.),TheCambridgeHandbookofSituatedCognition.Cambridge:CambridgeUniversityPress.
Koehler,D.J.(1991).Explanation,imagination,andconfidenceinjudgment.PsychologicalBulletin,110(3),499‐519.
Kotovsky,K.(1985).Whyaresomeproblemshard?EvidencefromTowerofHanoi.CognitivePsychology,17(2),248‐294.
Kotovsky,K.,&Simon,H.(1990).Whatmakessomeproblemsreallyhard:Explorationsintheproblemspaceofdifficulty.CognitivePsychology,22(2),143‐183.
Kühberger,A.,&Tanner,C.(2009).Riskychoiceframing:Taskversionsandacomparisonofprospecttheoryandfuzzy‐tracetheory.JournalofBehaviouralDecisionMaking,23(314‐329).
Lakoff,G.(2004).TheImportanceofCategorization.InB.Aarts,D.Denison,E.Keizer&G.Popova(Eds.),FuzzyGrammar:AReader.NewYork:OxfordUniversityPress.
Lord,C.G.,Ross,L.,&Lepper,M.R.(1979).BiasedAssimilationandAttitudePolarization:TheEffectsofPriorTheoriesonSubsequentlyConsideredEvidence.JournalofPersonalityandSocialPsychology,37,2098‐2109.
Mandel,D.R.(2001).Gain‐LossFramingandChoice:SeparatingOutcomeFormulationsfromDescriptorFormulations.OrganizationalBehaviourandHumanDecisionProcesses,85(1),56‐76.
Marmeche,E.,&Diderjean,A.(2001).Isgeneralizationconservative?Astudywithnovicesinchess.EuropeanJournalofCognitivePsychology,13,475‐491.
Marshall,C.C.,&Shipman,F.M.(2011).SocialMediaOwnership:UsingTwitterasaWindowontoCurrentAttitudesandBeliefsPaperpresentedattheCHI,Vancouver,BC,Canada.
Martino,B.D.,Kumaran,D.,Seymour,B.,&Dolan,R.J.(2006).Frames,Biases,andRationalDecision‐MakingintheHumanBrain.Science,313(5787),684‐687.
McNeil,B.J.,Pauker,S.G.,Sox,H.C.,&Tversky,A.(1982).Ontheelicitationofpreferencesforalternativetherapies.NewEnglandJournalofMedicine,,306,1259‐1262.
Medin,D.L.,&Rips,L.R.(2005).ConceptsandCategories:Memory,MeaningandMetaphysics.InK.J.M.Holyoak,R.G.(Ed.),TheCambridgeHandbookofThinkingandReasoning(pp.37‐72).Cambridge:CambridgeUniversityPress.
Mendoza,M.,Poblete,B.,&Castillo,C.(2010).TwitterUnderCrisis:CanwetrustwhatweRT?.PaperpresentedattheSOMA,Washington,DC,USA.
Mikels,J.A.,&Reed,A.E.(2009).MonetaryLossesDoNotLoomLargeinLaterLife:AgeDifferencesintheFramingEffect.TheJournalsofGerontology,SeriesB,64(4),457‐460.
193
Miu,A.C.,&Crişan,L.G.(2011).Cognitivereappraisalreducesthesusceptibilitytotheframingeffectineconomicdecisionmaking.PersonalityandIndividualDifferences,51(4),478‐482.
Moore,A.,Hayes,J.,&Wong,B.L.W.(2013).CartographicandCognitivePerspectivesonAmbulanceDispatchDisplays.InA.Moore&I.Drecki(Eds.),GeospatialVisualisation(pp.69‐88):SpringerBerlinHeidelberg.
Myers,D.G.(2002).Intuition:Itspowersandperils.NewHaven,CT:YaleUniversityPress.
News,B.(2009).RageAgainsttheMachinebeatXFactorwinnerincharts.Retrieved20thSeptember2011,2011,fromhttp://news.bbc.co.uk/1/hi/8423340.stm
Norman,D.A.(2002).TheDesignofEverydayThings.London:BasicBooks.Novemsky,N.,&Kahneman,D.(2005).TheBoundriesofLossAversion.Journal
ofMarketingResearch,42,119‐128.Patton,J.H.,Stanford,M.S.,&Barratt,E.S.(1995).Factorstructureofthebarratt
impulsivenessscale.JournalofClinicalPsychology,51(6),768‐774.Priem,J.,&Costello,K.L.(2010).HowandwhyscholarsciteonTwitter.Paper
presentedattheASIST,Pittsburgh,PA,USA.Reiter‐Palmon,R.,Illies,M.Y.,Cross,L.K.,Buboltz,C.,&Nimps,T.(2009).
Creativityanddomainspecificity:Theeffectoftasktypeonmultipleindexesofcreativeproblem‐solving.PsychologyofAesthetics,Creativity,andtheArts,3(2),73‐80.
Rosch,E.(2004).PrinciplesofCategorization.InB.Aarts,D.Denison,E.Keizer&G.Popova(Eds.),FuzzyGrammar:AReader.NewYork:OxfordUniversityPress.
Rosch,E.,&Mervis,C.B.(1975).Familyresemblances:Studiesintheinternalstructureofcategories.CognitivePsychology,7(4),573‐605.
Routh,D.K.,&King,K.M.(1972).Socialclassbiasinclinicaljudgment.JournalofConsultingandClinicalPsychology,38(2),202‐207.
Saletan,W.(2011).SpringtimeforTwitter.Retrieved20thSeptember2011,2011,fromhttp://www.slate.com/id/2299214/
Scribner,S.(1984).StudyingWorkingIntelligence.InB.Rogoff&J.Lave(Eds.),EverydayCognition:ItsDevelopmentinSocialContext(pp.9‐40).Cambridge,MA:HarvardUniversityPress.
Seo,M.‐G.,Goldfarb,B.,&Barrett,L.F.(2010).AffectandtheFramingEffectwithinIndividualsoverTime:RiskTakinginaDynamicInvestmentSimulation.AcademyofManagementJournal,53(2),411‐431.
Shafir,E.(1993).Choosingversusrejecting:Whysomeoptionsarebothbetterandworsethanothers.MemoryandCognition,21,546‐556.
Simon,H.(1957).ABehavioralModelofRationalChoiceModelsofMan,SocialandRational:MathematicalEssaysonRationalHumanBehaviorinaSocialSetting.NewYork:Wiley.
Simon,H.A.,&Reed,S.K.(1975).Modelingstrategyshiftsinaproblem‐solvingtask.CognitivePsychology,8(1),86‐97.
Sokol‐Hessner,P.,Hsu,M.,Curley,N.G.,Delgado,M.R.,Camerer,C.F.,&Phelps,E.A.(2009).Thinkinglikeatraderselectivelyreducesindividuals’
lossaversion.ProceedingsoftheNationalAcademyofSciencesoftheUnitedStatesofAmerica,106(13),5035‐5040.
194
Sperling,G.,&Dosher,B.A.(1986).Strategyoptimisationinhumaninformationprocessing.InK.R.Boff,L.Kaufmann&J.P.Thomas(Eds.),HandbookofPerceptionandhumanperformance:Volume1.Sensoryprocessesandperception(pp.2‐1‐2‐65).NewYork:JohnWileyandSons.
Stanovich,K.E.,&West,R.F.(2000).Individualdifferencesinreasoning:Implicationsfortherationality
debate.BehaviouralandBrainSciences,23,645‐665.Tanner,W.P.,&Swets,J.A.(1954).Adecision‐makingtheoryofvisualdetection.
PsychologicalReview,61,401‐409.Tom,S.M.,Fox,C.R.,Trepel,C.,&Poldrack,R.A.(2007).TheNeuralBasisofLoss
AversioninDecision‐MakingUnderRisk.Science,315(5811),515‐518.Tversky,A.,&Kahneman,D.(1981).TheFramingofDecisionsandthe
PsychologyofChoice.Science,211,453‐458.Volz,K.G.,&Cramon,D.Y.V.(2006).Whatneurosciencecantellaboutintuitive
processesinthecontextofperceptualdiscovery.JournalofCognitiveNeuroscience,18,2077‐2087.
Vosniadou,S.,&Ortony,A.(1989).Similarity,typicalityandcategorization.InS.O.Vosniadou,A.(Ed.),SimilarityAndAnalogicalReasoning(p.41).Cambridge:UniversityofCambridge.
Wason,P.C.(1966).Reasoning.InB.M.Foss(Ed.),NewHorizonsinPsychology.JournalofPersonalityandSocialPsychology,:Penguin.
Wigand,F.D.L.(2010).TwitterTakesWinginGovernment:Diffusion,Roles,andManagementPaperpresentedattheDG.O,Puebla,Mexico.
Zarnoth,P.,&Sniezek,J.A.(1996).TheSocialInfluenceofConfidenceinGroupDecisionMakingJournalofExperimentalSocialPsychology,33(4),345‐366.
195
AdditionalAcknowledgements
Ioweagreatdealofthanksforhavingmanagedtoreachthispoint.Itisowedtomanypeople.Andsomeinanimateobjects.Andafewabstractconcepts.Theygettheirduehere.
MumandDad–fordecidingtohaveasecondchildinthefirstplace,andforallthetime,money,patienceandendlesssupporttheyshowedevenasIblewdeadlines,wastedmoneyandseemedtobewastingmylife.Atnopointweretheyanythingbutwhollyenthusiastic,aqualitythatwassorelyneeded.
PeterJohnson–ForbeingexactlythekindofsupervisorIneededtoreachmypotential.TheamountI’vegrownreallyisremarkable,andit’smostlyyourdoing.
DaveSibley–Forendlesscupsofcoffee,pintsofbeer,theoreticalconversationsandgeekyinterest.AndGlastonbury.AndforbeingoneofthenicestmenIhaveevermetinmylife.
LouiseMissen–Foraconstantstreamofconversationsthroughavarietyofelectroniccommunicationmediumsthatkeptusbothamused.Well,meanyway.
JamesSutton–formaintainingcontactafterweleftNottinghamandputtingupwithmyshitatlongdistancewiththesameaplombheshowedincloserproximity.
JamesRosenberg,TimCoughlan,JohnCox,VickyShippandDuncan–ForendlessnightsinBacktoMineandotherdrinkingholesinandaroundBath.
DanCrick–ForAlcohol,mockeryandgenerosity.
GeorginaStubbings–Foralcohol,dancingandthesharededucationofagenerationofBathUnipsychologyfirstyears,maygodhavemercyontheirpoormangledminds.
BryanBrownlie–Forlongdistancesarcasmandhospitalityinfarflunglands.AlsoforbookendingtheperiodwhereIactuallyfiguredoutwhatIwasdoing.AndforbeingthefirstpersonwhotoldmeIwastotallygoingtodoaPhDbackwhenIwasindenial.
EsmeDark–Forfloorspace.Somuchfloorspaceforsleepingon.Oh,anddancingcrazytoo.Youcan’targuewithdancingcrazy.Andjustbeingthereprettymuchwhenever.
IanFairholm–ForprovidingtheotherhalfofmyPhDeducationbylettingmeteachothers.Itmightnotbeinthisthesisdirectly,butthisthesisisbetterforithavinghappened.Thestudentsmaynotbe,butwhocaresaboutthem?Andalsoforendlessprogrockreferences.
196
AgenerationofBathUniUndergraduatePsychologyStudents–Forbeingtaught,notreactingtoobadlyandbeingareliablesourceofparticipants.
Barristaseverywhere–Forcoffee.
Coffee–Yesitgetstwomentions,youhavenoideahowmuchIneededit.
TristanBrindle–forhandingmeaguitarafewweeksbeforeGlastonburyoneyearandtellingmetolearnsomething.Endedupkeepingmeremarkablysanethat.
Glastonbury–forbeingDisneylandforAdults.
Nath–Forbeingway,waybetterthanotherpeopleatguessingwordcounts.
Trains–BecausewithoutyouhowwouldIgetanywhere?
ThatonepresentationatIUI’09thatmademethink–Seriously,I’mnotsureI’dhavefinishedwithoutyouguys.Orfoundthistopic.Andyouweren’teventhatrelatedasatopic.Strangehowthemindworks.
85WellsRoad–ForbeingthebachelorpadIhopeditmight,whilstretainingridiculouslychinzdécor.
AliBagshaw–Generallyforservicesofenduringfriendship,butmostspecificallyforgettingmethroughthewritingupperiod.
Mysmallarmyofproofreaders–Mum,Dad,Dave,Ellie,Tess,Louise,Stacey,Kate,George,Putu:Youfuckingsavedmylife.
Dstl–Wait,I’memployable?
TheUKVettingAgency–Wait,I’mtrustworthy?
SalisburyPeople–Forgettingmethroughcorrections.Youmaynotknowwhatyoudid,butyouhelped.
KahnemanandTversky–Youguysareawesome,andI’veneverevenmetyou.
TheFineCityofBath–forhavingfarmoretoofferthanIimaginedpossiblewhenIfirstarrived.
WidcombeHill–forkeepingmefit.Although,seriously,whoputtheuniversityatthetopofyou?
StephenPayne–Ihatedprettymuchallmycorrections…butIsuspectthatdoingthemmademeabetterscientist.
EveryoneElse–Ican’tnameyouall,butyoumattered.Cheers.
197
Appendices
TheExperimentsinthisthesisrequiredagreatdealofmaterialsandtext.Ifthesewereprintedverbatiminthissectiontheywouldaddagreatdealoflargelyunnecessaryspace.
Consequently,asampleofallappropriatematerialsarepresentedhere.Somehavebeencompressedtosavespace:primarilyquestionnaireswhereaspacewouldotherwisehavebeenleftforasubject’sanswers:nowordshavebeenchangedorremoved.AfullarchiveofallthematerialsusedintheiroriginalformcanbefoundontheCDattachedtothiswork.Alternativelythefullarchivecanberequestedbyemailingtheauthorattimothyharrison@gmail.com.RawresultsdataisnotincludedontheCD,butmaybesupplieduponrequesttothesameaddress,ascanallrecordsofstatisticaltestsandanymaterialsthatmayhavebeenomittedbyaccident.
198
One:SolitaireStudy‐Pre‐StudyQuestionnaire
Questionnaire
Thepurposeifthisquestionnaireissimplytodocumentwhatyouunderstandaboutthegameofsolitairepriortoplayingit.Answertheopen‐endedquestionstotheextentthatyoufeelappropriate.
Part1
Answerthefollowingquestionsbycirclingtheappropriatenumberindicatinghowstronglyyoufeelbythefollowingscale:
1–StronglyAgree
2–Agree
3–Neutral,abitofboth,neitheragreenordisagree
4–Disagree
5–StronglyDisagree
Iknowhowtoplaysolitaire
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Ilikehavingadefiniteanswertothings.
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
IhavemethodsforsolvingproblemsthatIstickto.
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
I’mnotverygoodatstrategygames
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Improvisingisastrengthofmine
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Igetcangetlostinlittledetails
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Itrytoavoidambiguoussolutions
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Ienjoynewchallengesandproblems
199
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Iworkatthingsmethodically
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
I’mgoodatcardgamesingeneral
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Iconsidermyselftobegoodatsolitaire
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Isometimesforgetaboutthebigpicture
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Ifindithardtocomeupwithnewideaswhenfacedwithaproblem
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Part2
Whatdoyouunderstandaboutthegame‘Solitaire’?Giveabriefoverviewofthebasicpurposeofit,andthenecessaryequipment.
Howdotheconstituentpartsfittogether?Inwhatwaydotheyinteract?
Doyoupursueanyparticularstrategieswhenplayingagame?Canyouthinkofanytacticsyouemploy?Howdothesework,andwhydoyouusethem?
Doyouplayanyothercardgamesonaregularbasis?Howmanywouldyousayyouarefamiliarwith?Whatarethey?
Doyouplayanyothergames–ofanysort‐regularly?Ifso,whatarethey,andhowoftendoyouplay?Didyouplayanyinthepast?
Two:SolitaireStudy‐EndQuestions
Asafinalquestion,howdoyoufeelyouadaptedtothechangehalfwaythroughtheexperiment?Doyoufeelithelpedorhinderedyourgame?Didyouchangeanyofyourstrategies,ordidyoufitinintoexistingones?Ifso,how?
Doyouhaveanyothercommentsabouttheexperimentandwhatyoudidingeneral?
200
Three:CardGameStudy–Pre‐StudyQuestionnaire
Questionnaire
Thepurposeifthisquestionnaireissimplytodocumentwhatyouunderstandaboutcardgames.Answertheopen‐endedquestionstotheextentthatyoufeelappropriate.
Part1
Answerthefollowingquestionsbycirclingtheappropriatenumberindicatinghowstronglyyoufeelbythefollowingscale:
1–StronglyAgree
2–Agree
3–Neutral,abitofboth,neitheragreenordisagree
4–Disagree
5–StronglyDisagree
Ienjoylogicproblems
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
IregularlyengageinpuzzlesinthepapersuchasSudokuorCrosswords
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Iliketohaveanoverallplanwhenapproachingaproblem
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
I’mnotverygoodatstrategygames
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Iplaygamesrecreationally
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
I’mhappytoletotherstaketheinitiativeingames.
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Itrytoavoidambiguoussolutions
201
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
I’masystematicthinker
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
I’mquiteacompetitiveperson
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
I’mgoodatcardgamesingeneral
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
I’mhappytodevelopideasonthefly
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Isometimesforgetaboutthebigpicturewhenmakingadecision
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Ingames,agoodoffenseisoftenthebestdefence
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Idon’treallycareaboutwinningandlosing
1‐‐‐‐‐‐‐2‐‐‐‐‐‐3‐‐‐‐‐‐4‐‐‐‐‐‐5
Part2
AreyoufamiliarwiththeCardgames‘Uno’,Crazy8’s’orsomeothervariant?Describehowfamiliaryouconsideryourselftobe.
Doyouplayanyothercardgamesonaregularbasis?Howmanywouldyousayyouarefamiliarwith?Whatarethey?
Doyouplayanyothergames–ofanysort‐regularly?Ifso,whatarethey,andhowoftendoyouplay?Didyouplayanyinthepast?
202
Four:CardGameStudy–Post‐StudyQuestionnaire
Yourfinalscore:
Yourpositioninthegame:
Nowthatthegameisover,howdidyoufeelaboutplayingit?Whatstrategiesdidyouinitiallyintendtoemploy?Didthisalteratallasthegameprogressed?
Giventhechancetoplayagain,whatmightyoudodifferentlynexttime?Wouldyouusethesamestrategy?
Pleasealsofeelfreetoofferanyotherthoughtsorcommentsonyourplayingexperience.
203
Five:RiskTaking/ImpulsivenessQuestionnaire
1.
204
23. Sunbathing without sunscreen.
24. Bungee jumping off a tall bridge.
25. Piloting a small plane.
26. Walking home alone at night in an unsafe area of town
27. Moving to a city far away from your extended family.
28. Starting a new career in your mid-thirties.
29. Leaving your young children alone at home while running an errand
30. Not returning a wallet you found that contains $200
205
Six:F‐ScaleAuthoritarianismQuestionnaire
Please indicate how strongly you agree or disagree with each of the following statements, by making a note of the number of the appropriate response.
Strongly Disagree(1) - - Disagree(2) - - Not Sure(3) - - Agree(4) - - Strongly Agree(5)
1. Deviant sexual behaviour between consenting adults may be disagreeable but it should not be regarded as a crime. 2. No sane, normal, decent person would ever think of hurting a close friend or relative. 3. Many of the radical ideas of today will be the accepted practices of tomorrow. 4. People who want to imprison or whip sex criminals are themselves sick. 5. Obedience and respect for authority are the most important virtues children should learn. 6. Young people sometimes get rebellious ideas but as they grow up they ought to get over it and settle down. 7. It is all right for people to raise questions about even the most personal and private matters. 8. Insults to our honor are not always important enough to worry about. 9. Sex crimes such as rape and attacks on children deserve more than imprisonment; such criminals ought to be publicly whipped or worse. 10. Most honest people admit to themselves that they have sometimes hated their parents. 11. Racial profiling is a necessary method of identifying potential terrorists in today’s world. 12. Sex crimes such as rape and attacks on children are signs of mental illness and such persons belong in hospitals rather than prisons. 13. There is hardly anything lower than a person who does not feel great love, gratitude and respect for his parents. 14. What the young need most is strict discipline., rugged determination and the will to work and fight for family and country.
206
Seven:Introversion/ExtraversionQuestionnaire
Answer the following by choosing to what degree you agree, or disagree with each statement about yourself. Make a note of the appropriate response’s number as it applies to you. Strongly Disagree (1) - - Disagree (2) - - Agree (3) - - Strongly Agree (4)
1) I am a talkative person
2) I am generally quite lively
3) I enjoy meeting new people
4) I usually let myself go and enjoy myself at a lively party
5) I usually take the initiative in making new friends
6) I can easily put some life into a quiet social gathering.
7) I tend to keep in the background on social occasions
8) I enjoy socializing with people
9) I like plenty of bustle and excitement around me
10) I am mostly quiet when I am with other people
11) Other people would describe me as lively
12) I can get a party going
207
Eight:FestivalStudyInstructionsForNeutralandQuestionFrameconditions
Inthistaskyouwillbeassumingtheroleofsitemanagerforamid‐sizedmusicfestivalcalled‘Wish’.Youhavebeenhiredtooverseetheday‐to‐dayrunningofthesite,andinchargeofmakingsurethattheeventproceedsassmoothlyaspossible.
Inthecourseofthistaskyouwillbepresentedwithaseriesofhigh‐leveldecisionsabouteventsoccurringatthefestivalandthepotentialcoursesofactionavailable.Yourjobistomakeadecisionbetweenthechoicespresented,andattempttorunthebesteventpossible.Youwillbegivenanideaofhowsuccessfulyouarebeinginyourtaskbytwofeedbackmetrics:moneyandreputation.
‘Money’representstheamountofmoneythatthefestivalhasgained,savedorlost,asaresultofadecisionyouhavemade,comparedtotheprojectionsforifthefestivalhadbeenfunctioningnormallyornotdecisionwasmade.Itisgiveninpoundssterling,andthestartingpointiszero.
Reputationisameasureofpublicopinionregardingtheevent.Itispresentedintheformofapublicpoll,withabaselineapprovalof60%.
Forbothmeasurestherearenoguaranteesthatyoucanobtainapositiveoutcome.Attimesitmaybeacaseofminimizinglosses,orconverselyofmaximizinggains.Similarly,thereiselementofrandomchancetotheoutcomeofthesedecisions:youcanmaximizeyourchanceofsuccess,butcannotguaranteeit.Asinreallife,sometimesthingsmaygowrongevenifyoumakealltherightdecisions.Allthedecisionsyoumakehavethepotentialtoaffectwhichfuturedecisionsyouface,andhaveconsequenceslateron.Youshouldconsidertheimplicationsofthechoicesyoumakeasaresult.
Foreachdecisionyouwillbegivenachoicebetweenwhethertoattempttohandletheproblemon‐sitewiththepre‐existingfestivalresources,orwhethertocallinoutsidehelpsuchasthepolice,fireservicesetc.Youmustdecidewhichismostlikelytobeappropriateineachcase.
Havingmadethischoice,youwillthenalsobeabletochoosethelevelofconfidencethatyouhaveinyourdecision,where1representsverylittleconfidence,and7absoluteconfidence.
Foreachquestion,youwillhaveamaximumofthreeminutestobothreadandunderstandthesituation,andthenchooseyourresponse.Ifyoufailtochoosearesponseintheallottedtime,itwillbeassumedthatyoudidnothing,andyourscoresaffectedaccordingly.Itwillalwaysbebettertochooseanoptionthantonotpickone.
208
Nine:FestivalStudyOver‐FrameInstructions,External
Inthistaskyouwillbeassumingtheroleofsitemanagerforamid‐sizedmusicfestivalcalled‘Wish’.Youhavebeenhiredtooverseetheday‐to‐dayrunningofthesite,andinchargeofmakingsurethattheeventproceedsassmoothlyaspossible.
Inthecourseofthistaskyouwillbepresentedwithaseriesofhigh‐leveldecisionsabouteventsoccurringatthefestivalandthepotentialcoursesofactionavailable.Yourjobistomakeadecisionbetweenthechoicespresented,andattempttorunthebesteventpossible.Youwillbegivenanideaofhowsuccessfulyouarebeinginyourtaskbytwofeedbackmetrics:moneyandreputation.
‘Money’representstheamountofmoneythatthefestivalhasgained,savedorlost,asaresultofadecisionyouhavemade,comparedtotheprojectionsforifthefestivalhadbeenfunctioningnormallyornotdecisionwasmade.Itisgiveninpoundssterling,andthestartingpointiszero.
Reputationisameasureofpublicopinionregardingtheevent.Itispresentedintheformofapublicpoll,withabaselineapprovalof60%.
Forbothmeasurestherearenoguaranteesthatyoucanobtainapositiveoutcome.Attimesitmaybeacaseofminimizinglosses,orconverselyofmaximizinggains.Similarly,thereiselementofrandomchancetotheoutcomeofthesedecisions:youcanmaximizeyourchanceofsuccess,butcannotguaranteeit.Asinreallife,sometimesthingsmaygowrongevenifyoumakealltherightdecisions.Allthedecisionsyoumakehavethepotentialtoaffectwhichfuturedecisionsyouface,andhaveconsequenceslateron.Youshouldconsidertheimplicationsofthechoicesyoumakeasaresult.
Foreachdecisionyouwillbegivenachoicebetweenwhethertoattempttohandletheproblemon‐sitewiththepre‐existingfestivalresources,orwhethertocallinoutsidehelpsuchasthepolice,fireservicesetc.Youmustdecidewhichismostlikelytobeappropriateineachcase.
Externalsolutionsaremoreprofessionalandaccomplished,butmaybedelayedinarrivingorotherwiseheldup.Aproblemwillbesolved33%ofthetime.
Internalsolutionsarealwaysonsiteandreadytogo,butarelessprofessionalandable.Theywillfailtosolve2/3ofanyproblem
Havingmadethischoice,youwillthenalsobeabletochoosethelevelofconfidencethatyouhaveinyourdecision,where1representsverylittleconfidence,and7absoluteconfidence.
Foreachquestion,youwillhaveamaximumofthreeminutestobothreadandunderstandthesituation,andthenchooseyourresponse.Ifyoufailtochoosearesponseintheallottedtime,itwillbeassumedthatyoudidnothing,andyourscoresaffectedaccordingly.Itwillalwaysbebettertochooseanoptionthantonotpickone.
209
Ten:FestivalStudyOver‐FrameInstructions,Internal
Inthistaskyouwillbeassumingtheroleofsitemanagerforamid‐sizedmusicfestivalcalled‘Wish’.Youhavebeenhiredtooverseetheday‐to‐dayrunningofthesite,andinchargeofmakingsurethattheeventproceedsassmoothlyaspossible.
Inthecourseofthistaskyouwillbepresentedwithaseriesofhigh‐leveldecisionsabouteventsoccurringatthefestivalandthepotentialcoursesofactionavailable.Yourjobistomakeadecisionbetweenthechoicespresented,andattempttorunthebesteventpossible.Youwillbegivenanideaofhowsuccessfulyouarebeinginyourtaskbytwofeedbackmetrics:moneyandreputation.
‘Money’representstheamountofmoneythatthefestivalhasgained,savedorlost,asaresultofadecisionyouhavemade,comparedtotheprojectionsforifthefestivalhadbeenfunctioningnormallyornotdecisionwasmade.Itisgiveninpoundssterling,andthestartingpointiszero.
Reputationisameasureofpublicopinionregardingtheevent.Itispresentedintheformofapublicpoll,withabaselineapprovalof60%.
Forbothmeasurestherearenoguaranteesthatyoucanobtainapositiveoutcome.Attimesitmaybeacaseofminimizinglosses,orconverselyofmaximizinggains.Similarly,thereiselementofrandomchancetotheoutcomeofthesedecisions:youcanmaximizeyourchanceofsuccess,butcannotguaranteeit.Asinreallife,sometimesthingsmaygowrongevenifyoumakealltherightdecisions.Allthedecisionsyoumakehavethepotentialtoaffectwhichfuturedecisionsyouface,andhaveconsequenceslateron.Youshouldconsidertheimplicationsofthechoicesyoumakeasaresult.
Foreachdecisionyouwillbegivenachoicebetweenwhethertoattempttohandletheproblemon‐sitewiththepre‐existingfestivalresources,orwhethertocallinoutsidehelpsuchasthepolice,fireservicesetc.Youmustdecidewhichismostlikelytobeappropriateineachcase.
Externalsolutionsaremoreprofessionalandaccomplished,butmaybedelayedinarrivingorotherwiseheldup.Aproblemwillbeunsolved66%ofthetime.
Internalsolutionsarealwaysonsiteandreadytogo,butarelessprofessionalandable.Theywillguaranteetosolve1/3ofanyproblem.
Havingmadethischoice,youwillthenalsobeabletochoosethelevelofconfidencethatyouhaveinyourdecision,where1representsverylittleconfidence,and7absoluteconfidence.
Foreachquestion,youwillhaveamaximumofthreeminutestobothreadandunderstandthesituation,andthenchooseyourresponse.Ifyoufailtochoosearesponseintheallottedtime,itwillbeassumedthatyoudidnothing,andyourscoresaffectedaccordingly.Itwillalwaysbebettertochooseanoptionthantonotpickone.
210
Eleven:FestivalStudySampleDecision,NeutralandOver‐FrameConditions
Decision3
Itiscurrently1:30pm,anditwillsoonbetimeforoneofthemostpopularactsontheline‐uptotakethestage,aparticularlyhyped‐upandhighlyanticipatedactfromAmerica.Theywerebookedseveralmonthsago,butsincethattimehavehadsignificantchartsuccesswiththeirdebutalbumandasaresultalargecrowdisexpected,significantlymorethanwouldnormallybefoundatthistime.
Duetoothercommitmentshowever,thebandarelatearriving.Theyarecurrentlymakingtheirwaytowardsthesite,buthavebeenheldupbytrafficofstill‐arrivingfestivalgoers.Securityfearsthatthemoodofthecrowdcouldturnnastyifthereistoolongadelay.
ThePolicecanbecontactedtohelpwiththesemattersbyescortingvehiclesthroughtrafficifasked,butitisuncleariftheywillbeabletofindthebandintime.Sitemanagementthinkstheycanaltersomeoftheentryroutestoeasetrafficflowontheroadthatthebandareon,buttheywillstillbesignificantlydelayed.Theyareduetoplayanhour’smusic,andthenextactcannotbedelayedduetocontractsandtightscheduling.
211
Twelve:FestivalStudySampleDecision,QuestionFrameInternal
Decision3
Itiscurrently1:30pm,anditwillsoonbetimeforoneofthemostpopularactsontheline‐uptotakethestage,aparticularlyhyped‐upandhighlyanticipatedactfromAmerica.Theywerebookedseveralmonthsago,butsincethattimehavehadsignificantchartsuccesswiththeirdebutalbumandasaresultalargecrowdisexpected,significantlymorethanwouldnormallybefoundatthistime.
Duetoothercommitmentshowever,thebandarelatearriving.Theyarecurrentlymakingtheirwaytowardsthesite,buthavebeenheldupbytrafficofstill‐arrivingfestivalgoers.Securityfearsthatthemoodofthecrowdcouldturnnastyifthereistoolongadelay.
ThePolicecanbecontactedtohelpwiththesemattersbyescortingvehiclesthroughtrafficifasked,butitisuncleariftheywillbeabletofindthebandintime.Sitemanagementthinkstheycanaltersomeoftheentryroutestoeasetrafficflowontheroadthatthebandareon,buttheywillstillbesignificantlydelayed.Theyareduetoplayanhour’smusic,andthenextactcannotbedelayedduetocontractsandtightscheduling.
Ifthepoliceareaskedtohelp,thereisa66%chancethebandwillmisstheirplayingslot.
Ifthesitemanagementalterstheroadsystem,thebandwillbeintimetoplay20minutesoftheirset.
212
Thirteen:FestivalStudySampleDecision,QuestionFrameExternal
Decision3
Itiscurrently1:30pm,anditwillsoonbetimeforoneofthemostpopularactsontheline‐uptotakethestage,aparticularlyhyped‐upandhighlyanticipatedactfromAmerica.Theywerebookedseveralmonthsago,butsincethattimehavehadsignificantchartsuccesswiththeirdebutalbumandasaresultalargecrowdisexpected,significantlymorethanwouldnormallybefoundatthistime.
Duetoothercommitmentshowever,thebandarelatearriving.Theyarecurrentlymakingtheirwaytowardsthesite,buthavebeenheldupbytrafficofstill‐arrivingfestivalgoers.Securityfearsthatthemoodofthecrowdcouldturnnastyifthereistoolongadelay.
ThePolicecanbecontactedtohelpwiththesemattersbyescortingvehiclesthroughtrafficifasked,butitisuncleariftheywillbeabletofindthebandintime.Sitemanagementthinkstheycanaltersomeoftheentryroutestoeasetrafficflowontheroadthatthebandareon,buttheywillstillbesignificantlydelayed.Theyareduetoplayanhour’smusic,andthenextactcannotbedelayedduetocontractsandtightscheduling.
Ifthepoliceareaskedtohelp,thereisa33%chancethebandwillmakethegigontime.
Ifthesitemanagementalterstheroadsystem,thesetwillbe40minutesshorter.
213
Fourteen:FestivalStudyQualitativeFeedbackPatternandValues
Money Reputation Cases
Question Decision Internal External Internal External Internal External
1 WaterPipe SmallGain SmallGain SmallGain SmallGain 0 0
2 FirstOutbreak SmallGain SmallLoss SmallGain SmallLoss 0 0
3 BandDelay MediumLoss MediumGain SlightLoss SlightGain 15 15
4 MedicalEvac LargeGain LargeLoss MediumGain MediumLoss NoChange SmallRise
5 CaseSpike MediumLoss MediumGain MediumLoss MediumGain SmallRise MediumRise
6 Rumours Mediumgain MediumLoss SmallGain SmallLoss VeryLargeRise VeryLargeRise
7 FoodPoisoning VeryLargeLoss LargeLoss MediumLoss MediumGain MediumRise SmallRise
8 SmallRiot LargeLoss VeryLargeLoss MediumGain MediumLoss SmallRise MediumRise
9 LeavingEarly MediumLoss MediumGain LargeLoss LargeGain MediumRise SmallRise
10 Diagnosis MediumGain MediumLoss LargeGain LargeLoss MediumRise LargeRise
11 Treatment MediumLoss MediumGain SmallLoss SmallGain VeryLargeRise VeryLargeRise
214
Fifteen:FestivalStudyNumericFeedbackPatternandValues
Money Reputation Cases
Question Decision Internal External Internal External Internal External
1 WaterPipe 215 215 1 1 0 0
2 FirstOutbreak 765 ‐765 1 ‐1 0 0
3 BandDelay ‐1892 1892 ‐1 1 15 15
4 MedicalEvac 4004 ‐4004 2 ‐2 0 2
5 CaseSpike ‐872 872 ‐2 2 3 3
6 Rumours 1756 ‐1756 1 ‐1 20 20
7 FoodPoisoning ‐15,472 ‐8765 ‐2 2 4 2
8 SmallRiot ‐8801 ‐15563 2 ‐2 3 5
9 LeavingEarly ‐4037 4037 ‐3 3 4 2
10 Diagnosis 2356 ‐2356 3 ‐3 8 12
11 Treatment 2405 2405 ‐1 1 23 19
215
Sixteen:FestivalStudyDecisionPageScreenshot
Seventeen:FestivalStudyFeedbackPageScreenshot
216
Eighteen:HerbalStudyAlternativeTherapiesattitudequestionnaire
Thetaskyouareabouttoundertakeinvolvesmakingpurchasingdecisionsinthecontextofalternativemedicine.Pleasetakeamomenttoanswerthefollowingquestionsonthattopic.
Answerthefollowingonthesheetprovidedonascaleof1‐5,with5being‘stronglyagree’,1being‘Stronglydisagree’and3as‘Neutral/Unsure’
1)Alternativemedicinescanproviderealtherapeuticeffects
2)Alternativemedicinecontainssystemsequalinvaluetothatofconventionalmedicine.
3)ConventionalmedicineisbetterthanAlternativemedicine.
4)Alternativemedicinehasnomerit
5)Peopleshouldbefreetousealternativemedicine,iftheychooseto
6)AlternativemedicineshouldbeavailableontheNHS
7)Thereisalotwecanpotentiallylearnfromalternatemedicinalsystems
8)Alternativemedicineisapoorsubstituteforconventionalmedicine
9)Alternativemedicineisnobetterthanafraud.
10)Peoplewhosellalternativemedicinesaredoingsomethingmorallyobjectionable.
11)Alternativemedicinecanbeusedinconjunctionwithconventionalmedicineforadditionaltherapeuticbenefit.
Pleaseanswerthefollowingsimplyyes/no:
12)Haveyou,oranyoneyouarecloseto,madeuseofanalternativetherapy?
13)DoyouthinkAlternativeTherapiescanbeeffective?
14)Wouldyourecommendanyalternativetherapiestoafriend?
217
Nineteen:HerbalStudySocialMediaQuestionnaire
Pleaseanswerthefollowingquestionsyes/no,orotherwiseasindicated.
1) Wouldyousayyouarefamiliarwithwhat‘socialmediais?’
2) Doyouhavea‘socialmediaaccount’ofanytype(twiter,facebook,
linkedinetc)?
3) Ifso,howmanydifferentonesdoyouhave(approximately)
4) Howoftenwouldyousayyoucheckasocialmediasiteofanytype?
(Severaltimesaday/Daily/afewtimesaweek/onceaweek/rarely)
5) AreyoufamiliarwithTwitterspecifically(whatitdoes,howitworks)?
6) Doyouhaveatwitteraccount?
7) Ifyes,howoftenwouldyousayyoucheckyourtwitter?
(Severaltimesaday/Daily/afewtimesaweek/onceaweek/rarely)
8) Areyoucomfortablewithreadingatwitter‘feed’asasourceofnewsetc?
9) Pleaserankthefollowingintermsofwhereyoufeelyougetthemajorityofyournewsfrom:Online(newssites)Online(Socialmediaetc)TVNewspapersWordofmouthRadio
10) Overall,howimportantwouldyousaythatsocialmediaistoyoureverydaylife–personal,professionalorotherwise?(Essential/Veryimportant/Useful/Sortofimportant/Unimportant)
218
Twenty:HerbalStudyInstructions
Inthistaskyouwillbetakingontheroleofabuyerforamediumsizedgroupofco‐operativeshops.Althoughyourdutiescoverarangeofproducts,thetaskyouwillbeaskedtoundertakewillfocusonyourpurchasingofalternativemedicinesandspecificallyherbalremedies.
Alternativemedicines–remediesthatexistoutsidetheframeworkofconventionalscientificprocess–havebeengrowinginpopularitygreatlyoverthelastdecade,andmanypeoplenowusethemforeverythingfromskinconditionstotreatingdepression.
Asabuyerforacooperative,youholdaresponsibilityforalargenumberofprivatelyheldandoperatedstores,whobuyinbulkasagroupandthenselltheseremediesaspartoftheirbusiness.Itisyourjobtobalancethevariousfactorsthatmaycomeintoplayandobtainproductsforwhichthereishighdemand,acompetitiveprice,canbereliablysupplied,etc.Notethatitisnotyourjobtojudgeefficacyoftheproducts,butratheraimtomeetthedemandthatexists!
Overthelastfewyearsyouhavedevelopedlinkswithtwocompanieswhoareindirectcompetitionwitheachotherandsupplymanyofthesameproducts.Itisyourtasktochoosewhichcompanyyouwillorderyourproductsoffforthatparticularmonth.Thetwocompaniesare:
AstorRemedies–acompanylocatedinthesmallcentralAsiancountryofAploniathatboastslocalproductionand‘ancientculturalwisdom’asthebasisforitsproducts.
Quetia–acompanybasedinthefirst‐worldantipodiancountryofYokoviawhichclaimstomakethemostofmodernproductiontechniquesfortheseoldideas.
Foreachdecisionyouwillbegiventhesituationasitisunderstoodbythemedia,andalsoasnapshotofyourtwitterfeedonthatday.Youhavechosenarangeofpeopleto‘follow’,eachofwhomisapersonofinterestinthearea(shopkeeper,journalist,aviduseretc)andaresharingtheirthoughtsontheissuesoftheday.
Inadditiontochoosingwhichcompanyyouwishtoorderfrom,youwillalsobeaskedtorateyourconfidenceinthisdecision,from1(littleornoconfidence)to7(highorabsoluteconfidence).Aftereachdecisionyouwillbepresentedwithfeedbackonhowyourchoicehasimpactedsalesandshopkeeperconfidence.Aftereachdecisionhasbeenmade,youwillthenreceivefeedbackaboutyourperformance,intheformofanoverallsalesreport,andalsothegeneralopinionsoftheretailersyouserve.
Eachoftheseareas(tweets,decision,choicemaking)arehiddenwhilsttheothersarebeingviewed.Youmay,however,switchbackandforthbetweenthesectionsforaslongasyoulike,andasmanytimesasyoulike.
Allthedecisionsyoumakehavethepotentialtoaffectthefuturedecisionsyoufaceandhaveconsequenceslateron,andshouldthereforebeapproachedtactically.Additionally,thereisanelementofrandomchancetotheoutcomeof
219
thesedecisionsandhowwelltheygo:youcanmaximizeyourchanceofsuccess,butcannotguaranteeit.Asinreallife,sometimesthingsmaygowrongevenifyoumakealltherightdecisions.
220
Twenty‐One:HerbalStudyFeedbackPatternandValues
Astor Quetia
Question Decision Correct Money Opinion Money Opinion
1 FirstChoice Either SmallGain SmallGain SmallGain SmallGain
2 MiracloShipping Astor SmallGain MediumGain SmallLoss MediumLoss
3 Childlabour Quetia MediumLoss SmallLoss MediumGain SmallGain
4 Yokregulating Astor MediumGain SmallGain MediumLoss SmallLoss
5 Glass Quetia SmallLoss SmallGain SmallGain MediumGain
6 InternalQ Astor MediumGain MediumGain MediumLoss SmallGain
7 Takeover Quetia SmallGain MediumLoss LargeGain MediumGain
8 MP Astor LargeGain SmallGain SmallGain SmallLoss
9 Container Quetia MediumLoss SmallLoss MediumGain SmallGain
10 GPletters Astor MediumLoss SmallLoss LargeLoss MediumLoss
221
Astor Quetia
Question Decision Correct Money Opinion Money Opinion
11 Productsreturned Quetia LargeLoss MediumLoss MediumLoss SmallLoss
12 PRinitiatives Astor SmallGain MediumGain SmallLoss SmallGain
13 Parliament Quetia SmallLoss SmallGain SmallGain MediumGain
14 Offers Astor SmallLoss SmallGain SmallGain SmallLoss
15 Enquirydue Quetia SmallGain SmallLoss SmallLoss SmallGain
16 Finaldecision Either MediumGain MediumGain MediumGain MediumGain
222
Twenty‐Two:HerbalStudysampledecision
Decision2
Inthelastmonth,therehasbeenaflurryofinterestinanewproductderivedfromaspecialtypeofflower.Itissaidtomakeapersonmoreconsistentlyawareandawake,aswellasboostingintelligence.Students,doctorsandotherhigh‐intensitygroupshavebeenshowingadistinctinterestinthesubstance,knownas‘Miraclo’.
Bothcompaniesyouareengagedwithhavestartedtoputverysimilarformsofthisproductintoproduction,andareofferingitasapurchaseoptionthismonth.ThemainconcernatthispointisintheviabilityofgettingthisproductdeliveredandontoshelvesintimetocapitalizeonthecurrentwaveofpublicitybeinggeneratedbymagazinesandTVshows.
Astor,locatedinAploniaisclosertoyourshopsbyseveralhundredmilesandexpecttodeliveryquickly.However,muchofitsdistributionnetworkreliesheavilyontrucksandroadtransportwhichareknowntobeoccasionallyunreliablethankstothecountry’sinfrastructureandgovernmentcorruptionwhichsometimesholdupshipmentssignificantlyandunpredictably.
Quetiadonothavethisissue,asthecountryisafullfunctioningfirst‐worldeconomy,howevertheyshiptheirproductsprimarilybyseainordertokeepdowncostsandremaincompetitive.Theirdeliverytimeisguaranteed,butatanon‐triviallylongertimeframethantheoptimalestimatesfromAstor.
223
Twenty‐Three:HerbalStudySampleTweets,NeutralDecision2
Twenty‐Six:HerbalStudySampleTweets,Pro‐AstorDecision2
224
Twenty‐Four:HerbalStudySampleTweets,Pro‐QuetiaDecision2
Twenty‐Eight:HerbalStudyInstructionPage
Screenshot
225
Twenty‐Five:HerbalStudySampleFeedbackScreenshot
226
Twenty‐Six:HerbalStudyEventText
‐‐‐‐BreakingNews‐‐‐‐
Overthelastthreedaysagreatnumberofpeoplehavebeentakenillacrossthecountrywithseveredigestiveandrespiratoryproblems.CaseshavebeenreportedinareasasfarapartasExeterintheSouthwest,NewcastleintheNortheastandNottinghaminthecenterofthecountry.
Althoughitwasnotimmediatelyclearwhattheproblemwas,consultationwithpatientsfamiliesandcoordinationbetweenvarioushealthauthoritieshaveidentifiedthat90%ofcasescanbedirectlycorrelatedwiththevictimshavingtakenMiraclointhelastfewdays.Subsequenttestsonthebottlesofthissubstancefoundatthehousesofpatientshaverevealedconsistentlevelsofindustrialfertilizer,wasteproductsandratpoison.BottlesofbothAstorandQuetiamarkedproductshavebeendiscoveredtobecontaminatedasaresult.Strangely,thetwoshowdifferenttypesofcontaminationbetweenthetwocompanies,butconsistencywithinagivencompany’sproduct.
Asaresultofthisdiscovery,anemergencyhealthywarninghasbeenissuedbythegovernment,withpeopleadvisedtodisposeofalloftheseproductstheyhave,regardlessofmanufacturerorlocationtheywereboughtfrom.Allsalesofthesubstancehavebeenbanned,andstockremovedfromshelvesacrossthecountry.Thegovernmenthasalsoruledoutthepossibilityofthisbeingaterroristormaliciousattackofanysort.
Detailsarestilltricklingthrough,butinanefforttotrackthesourceoftheproblem,bothAstorandQuetiaproductshavebeenidentifiedascausalfactors.Bothcompaniesareinsistingthatthisistheresultoffraudulentknock‐offproductsthattheyarenotresponsiblefor,althoughtherearesomesuggestionsthattherewascontaminationwithinregularshipmentstoo.
AstorFramedtext:
Currentlydoctorssaythattheyhavebeenabletosave1/3ofthepatientsthattookAstorproducts.However,thereisa66%chancethatthepatientswhotookQuetiaproductswillsufferseverlong‐termhealthissuesanddisabilities.
QuetiaFramedtext:
Currentlydoctorssaythattheyhavebeenabletosave1/3ofthepatientsthattookQuetiaproducts.However,thereisa66%chancethatthepatientswhotookAstorproductswillsufferseverlong‐termhealthissuesanddisabilities.
227