Download - Collective Intelligence and Three Aspects of Planning in ...

Transcript

CollectiveIntelligenceandThreeAspectsofPlanninginOrganizations:ANASAExampleDorritBillman,SanJoseStateUniversityResearchFoundation[[email protected]]&MichaelFeary,NASAAmes[[email protected]]CollectiveIntelligenceandPlanning.

Weconstrue“collectiveintelligence”(CI)intermsofa)activityandb)outcome,ratherthanintermsofthesupportingtechnology:collectiveintelligenceisa)thegeneration,apprehension,anduseofdistributedinformationthroughcollaborativeprocessesandb)theknowledgeorproductsresultingfromsuchactivity.

Ourperspectiveoncollectiveintelligenceconcernsplanningwithinanorganization.By“planning”—theprocess,wemeanidentifyingeventsandschedulingthemintime,withtheintenttoguideoranticipatefutureactions.Byplan—theproduct,wemeananexternallyrepresentedspecificationofa(partially)orderedsetofactivities.Inorganizations,planningdependsoncollectiveintelligenceandplansinturnprovideinputtocollectiveintelligence.Whilestudyofplanningbypeopleandcomputershasalongandvariedtradition(e.g.Wilensky,1983;Fikes&Nilsson,1971),designingeffectivetoolstosupportcomplexplanningbygroupsofpeopleremainsaverydifficultproblem(Butler,Zhang,Esposito,Bahrami,Hebron,&Kieras,2007).

OurcoreexamplecomesfromthedomainofplanningbytheADCOcontrollergroup(AttitudeDeterminationandControl),whichispartofNASA’smissioncontroloftheInternationalSpaceStation.ThisgroupplansthedetailedsequenceofactivitiesneededtopositiontheISS,whilecollaboratingacrossgroupstoensuretheappropriatenessoftheplannedactivitiesinthewidercontextoftheISSmission.Inaddition,weareinformedbyNASAAmesresearchandapplicationdevelopmentforseveralotherplanningdomains,includingplanningforMarsRoverscientificactivityandschedulingactivitiesinabedrestfacility(McCurdy,Luowise,Marquez,&Li,2009).

TemporalStructureofPlanning

Planning/Schedulingisorganizedaroundtemporalstructure,anditstemporalstructureisdistinctiveandcomplex.Further,fromatechnologypointofview,CItoolsthatworkeffectivelyfordomainsnotorganizedaroundtemporalstructure(inmanyexamplesofCIfrombookmarkingtowiki‐writing)maybeunsatisfactoryfordomainsprimarilystructuredbytime.Nevertheless,planningtypicallyisandoughttobeacollectiveintelligenceactivity.Indeed,theneedforcomplexplanningandschedulingprimarilyarisesforcomplexsociotechnologicalsystems.Suchsystems

demandparticipationofmultiple,coordinatedplayers.Theneededparticipationwillvaryinhowtightlycoupledboththetimingandtheintentionsoftheplayersneedtobe,butcollectivelydistributingandconsolidatinginformationiscritical.

Planninghasthreeimportanttemporalprofiles,whichwelabelfuture‐focused,present‐focused,andpast‐focused.Withineachofthese,therelevantactivitycanvaryintheextentandnatureofcollaboration,thatis,intheorganizationofcollaborativeintelligence.Recognitionofthesethreeprofilesisimportantbecauseeachisoftenanalyzedandsupportedinrelativeisolationfromtheothers,yetworktypicallyinvolvesswitchingamongandcoordinatingthesetemporalprofiles.

Future‐focusedprofile.Whenbuildingaplan,usersarefocusedonthefuture.Theactivityofplanning,distinctfromtheplannedactivity,isitselfoftenacomplex,andcollaborative,process.Incollaborativeplanning,decision‐makingbyoneindividualorgroupdependsoninformationfromothergroupsorindividualsandinturninformationfromthefirstgroupisneededasinputbyothers.Theremaybeabroad“upstream‐downstream”directionofdependencies,andindeedproceduresaredevelopedtotrytoregularizeandreduceinterdependenciesininformationflow.Nevertheless,therefrequentlyisconsiderableinteractivityinmakingthedecisionsthatarereifiedintheplan.Interactionmaybeformalizedinpart,asinapprovalprotocols,butitalsoflowsviainformalrequestsandoffers.Informalcommunicationstypicallyenablethesuccessofformalcommunications.Afuture‐focusedprofileofplanninginvolvescollectiveintelligencetotheextentthatconstructionoftheplaniscollaborativeanddrawsonavarietyofinformationsources.

Present‐focusedprofile.Whenexecutingaplan,usersarefocusedonthepresent.Aplanissuccessfultothedegreeitsupportsexecution;planninginsociotechnicalsystemsisdrivenbytheneedtocoordinateexecutionofcomplextasks.Fluentandaccurateexecutionofsufficientlycomplextasksisonlypossiblewheninformationhasbeengathered,understood,andappliedinadvance,sothatthebulkofdecisionshavebeen“precompiled”inaplan,leavingthesmallestpossiblesettobeaddressed“intheheatofaction.”Executionoftenrequirescoordinatedactionbymultipleindividualsinmultiplegroups.Asuccessfulcollaborativeplanmustsupporttheparticipatinggroupsincarryingouttheactionswithintheirresponsibilityanditmustensurethattheactivitiesacrosscontributinggroupsareappropriatelycoordinated.Apresent‐focusedprofileofplanninginvolvescollectiveintelligencetotheextentthatexecutionoftheplanned‐foractivityiscollaborative.

Past‐focusedprofile.Aftertheaction,aplanbecomeshistory.Itprovidesarecordofintensionsand,ifupdatedorrevisedwithwhateventsactuallytookplace,ahistoricalrecord.Whenanalyzingenactedplans,usersarefocusedonthepast.Arecordofactualevents,intendedevents,andtheircomparisonisverypowerfulinformation.Oldplanscanbeusedastemplatestoefficientlygeneratesimilarnewplansandtheycanprovidetheinformationtoidentifypastproblemsorinefficienciesanddesignimprovedplansforthefuture.However,itisoftenverydifficultforanorganizationtomakeeffectiveuseofenactedplanstoguide

subsequentiterationsofplanningandacting.Apast‐focusedprofileofplanninginvolvescollectiveintelligencetotheextentthatreuseandanalysisofoldplansiscollaborative.

AnExample:PlanningintheADCOmissioncontrolgroup

WeusethesethreetemporalprofilestosummarizetheplanningactivitiesinADCO,anddescribesomeofthechallengesindesigningsupportfortheactivities.Ourfocushasbeenonthefuture‐focusedprocessofplanning,butthispromptedbothobservationsandquestionsaboutthepresent‐andpast‐focus.Inaddition,wementioncontrastingcasesofplanningtopointoutthevarietyofplanningneeds.

Future‐focus.InADCO,planningisdonebythesamegroupofindividualsasthosewhowillbeexecutingtheplan.AnADCOplangetshighlevelinputfromthemissionplanners.ADCOandRussiancounterpartsthenplanthedetailsoftherequiredmanuovers.PlansarepassedbackandforthbetweenADCOandRussiangroupsthroughaseriesofformalapprovalsteps.ADCOdistributesitsdevelopingplantootherdisciplines,suchasthoseresponsibleforpower,whoneedtoknowtheorientationoftheISSfortheirownplanning.Inturn,theorientationsspecifiedinthedevelopingplanmaybeadjustediftheyundulyrestricttheabilityofothergroupstomeettheirownrequirements.

Planningiscomplicatedbythenestedstructureofactivityandofplans.Activitiesdefinedbyhigher‐levelgoalsandrunningoverlongertimespanstypicallyhaveinstrumental,shorteractivitiesnestedwithinthem.Planninghigherlevelactivitiestypicallyprecedesspecificationoflowerlevelactivitiesintendedtoaccomplishthehigherlevelunit,andofteniscarriedoutbydifferentgroupsorindividualsthanthoseplanningthelower,moretacticalactions.ForADCOtheplanningprocessitselfisstrictlyscheduled,with30‐day,7‐day,3‐day,and1‐dayplanreviewsandincreasingdemandthatalldetailsarespecifiedandunchangedasthetimeofexecutionapproaches.Planpropagationisawkward,asADCOentersthesameinformationinatleast3products.Further,ADCOcannottakeahighlevelplandevelopedformissionlevelplanningasinputandusethistoseedtheirownmoredetailedplandevelopment.

Present‐focus.DuringmissionexecutionmanyMissionControlgroupsworktogethertomonitorandcontrolthesubsystemsofthespacestation.Inaddition,coordinationwithRussiancounterpartsisrequired.Thetechnicalandsocialcomplexityrequiresboththatmanytasksaredoneinadvance(eitherdirectlybyhumansoroff‐loadedintocomputerizedprocedures).Fornovelorriskyactivities,ADCOalsohasbackupplans,inwhichsolutionsapplicabletopotentialproblemsituationsareprecomputedandmayactuallybeuploadedandavailableforuseontheISS.Duringexecution,theADCOoperatorincontrolisco‐locatedwiththecontrollingoperatorsfromothermissioncontroldisciplines.Thereisimportantinformalcommunicationamongdisciplines,suchasalertingotheroperatorsthatananticipatedandunproblematicwarningmessagewillbeappearingshortly.Inotherdomains,suchasfacilityscheduling,real‐timeexecutionoftheplan(e.g.puttinga

subjectinthecorrectroom)mayberelativelysimple,andrequirerelativelylittlecoordination.

ForADCO,thesamecollectionofinternationalpartnersanddisciplineswithinMissionControlarecollaborativelyinvolvedinbothplanningandexecution.Incontrast,MarsRoveractivitiesareplannedbyonegroupofscientists,butadifferent,engineeringgroupmonitorstheRover’sactivities,whicharedesignedtobeexecutedautonomously.

Past‐focus.ADCOheavilyreusesoldplans,toactastemplatesfornewplans.Thereareatleasttwotypesofimpactthatreusehas.Reuseincreasesefficiencyofplanning,becauseadaptingapriorplantypicallyisasmallpartoftheeffortofcreatinganewone.Thisisclearlytrueatthelevelofthemechanicalconstructionoftheplan,asthisrequirestediousanderror‐vulnerabledataentry.ReuseisalsorelatedtotheconservativeplanningstanceofADCO.ADCOplannersstatethattheyareprimarilyconcernedwithgettingasufficientlygoodplan,thatiscorrectlycommunicatedandimplemented,ratherthanbuildingthebestplan.Thisisencouragedbyandencouragesthereuseofoldplans:notonlydoesitreduceeffortatthemechanicallevel,butitisbelievedtoreduceriskattheanalyticallevel.Doingsomethingverysimilartowhathasbeendonepreviouslyislikelytobesafer.Whilethisapproachisguidedbyofficialflightrulesaboutwhatactionsaresanctioned,italsoisexpressedininformalprocesses.

ADCOupdatestheactiveplantoshowhowitwasmodifiedinexecutionandsavesan“as‐flown”plan.ADCOoperatorsrecognizethevalueoftheseforidentifyingplaceswhereproceduresmightchangeorwherethesituationwasnotasanticipated;wedidnotseeorinvestigatewhenorbywhomretrospectiveanalysisissystematicallydone.Retrospectiveanalysisofplansremainsaninterestingopenissue.Forhighrisk,safety‐criticaldomains,thisretrospectiveaspectofplanningmaybewherecollectiveintelligence,distributedcollaboration,andsocialnetworkingtechnologiesmaybemostvaluable.

Supportforplanning.

Formanycomplexsociotechnicalsystems,planningseemstorequireexplicitcoordination;certainly,inexecutingaplantheactivitiesofdifferentactorsmustbetightlycoupled.However,distributingtheneededplanninginformationcanbeveryburdensomeanderrorprone,becausedifferentgroupsneeddifferentcollectionsofinformation,updatedorkeptcurrentondifferenttimecycles.Further,theinformationneededtoformsuccessfulplansisoftenhighlydistributed,andwhilefeedbackaboutthesuccessofpriorplansmayexist,itmaynotbeavailabletothoseinapositionofusingthistoimproveplansortodetectandresolveotherproblemsinthesystem(Weick,1995).

Toolstosupportvariousaspectsofplanninghavebeendeveloped,andcanprovideahugebenefittotheindividualsworkingonthataspect.Tobetractable,mostsolutionsaddressaquiteboundedsliceofwork,isolatingitfromthelargercontext.

Prospectiveplanningtakesplaceovermultiple,nestedcyclesofdecisionmaking.Thisbuildsaplanthatspecifiesactivitiesofdifferentgranularity.“Subplans”mayspecifymultipleparallelactivitiesbydifferentgroupsandindividuals,aswellassequential,nestedactionsbyasingleactor.Planningproducesvaluable,sharable,external,representations:inadditiontoprospectiveuse,planssupportretrospectiveassessmentandalsoactioninthepresent.Viewingplanninginalargercontext–bothtemporalandorganizational‐‐enablesnoticingwhatonedoesnotknowandgeneratingmoresystemicandeffectivesolutions.

Viewingaproblemasoneofcollectiveintelligenceinvitesthinkingaboutthelargerorganizationalcontext.Manyapproachestosupportingcollectiveintelligencedonotsupportexecutionofhighlycontingentactions,distributedacrossmanyplayers,andhenceprovideincompletesupportforplanning.However,CItechnologymaybehelpfulinmanagingtheprocessesofgatheringinformationfordecisionmakinginplanningandofdistributingplaninformationtovariousactors,neededonvarioustimecycles.

References

Butler,K.E.,Zhang,J.,Esposito,C.,Bahrami,A.,Hebron,R.,&Kieras,D.(2007).Work­CenteredDesign:ACaseStudyofaMixed­InitiativeScheduler.CHI2007Proceedings,NewYork:ACM.Pp747‐756.

Fikes,R.E.,&Nilsson,N.J.(1971).STRIPS:Anewapproachtotheapplicationoftheoremprovingtoproblemsolving.ArtificialIntelligence,2,189–208.McCurdy,M.,Ludowise,M.,Marquez,J,andLi,J.(2009)SpaceHumanFactorsEngineeringReport:CrewSchedulingLessonsLearned.NASAAmesinternalreport.

Weick,K.E.(1995)SensemakinginOrganizations.ThousandOaks,CA:SagePublishing.

Wilensky,R.(1983)PlanningandUnderstanding:AComputationalApproachtoHumanReasoning.Reading,MA:Addison‐WesleyPublishingCompany.