Envisioning the future of personalization through personal informatics: A user study

DOI: 10.4018/IJMHCI.2018010104



Copyright©2018,IGIGlobal.CopyingordistributinginprintorelectronicformswithoutwrittenpermissionofIGIGlobalisprohibited. 

Envisioning the Future of Personalization

Through Personal Informatics:

A User Study

Federica Cena, Computer Science Department, University of Torino, Italy Amon Rapp, Computer Science Department, University of Torino, Italy Silvia Likavec, Computer Science Department, University of Torino, Italy Alessandro Marcengo, TIM, Torino, Italy

ABSTRACT Inrecentyears,UserModeling(UM)sceneryischanging.Withtherecentadvancementsinwearable andmobiletechnologies,theamountandtypeofdatathatcanbegatheredaboutusersandemployed tobuildUserModelsisrapidlyexpanding.UMcannowbeenrichedwithdataregardingdifferent aspectsofpeople’sdailylivesandislikelytodelivernovelpersonalizedservices.Allthesechanges bringforthnewresearchquestionsaboutthekindsofserviceswhichcouldbeimproved,whichof themwouldbethemostuseful,thewaysofconveyingeffectivelynewformsofrecommendations, andhowuserswouldperceivethem.Inthispapertheauthorstriedtofindanswerstosomeofthese questionsbyexploitinganovelpersonalizedsystemtoconductaqualitativeuserstudy,withtheaim tounderstandusers’needsandexpectationsw.r.t.personalizationenabledbythepresenceofwearable andmobiletechnologies. KEywoRdS

Personal Informatics, Personalization, Quantified Self, Recommendations, Self-Tracking, User Model

INTRodUCTIoN Withtherecentdevelopmentofwearableandmobiletechnologies,theamountandtypeofdatathat canbegatheredaboutusersisrapidlyincreasing(Rappetal.,2015).Therearemanyaspectswhich contributetothistrend.Firstofall,theparadigmofInternetofThings(IoT)(Gubbietal.,2013) digitallyconnectseverydayobjectsintherealworldmakingdatapervasive.IoTbringstolifeWeiser’s visionofubiquitouscomputing(Weiser,1998)whichtriestobringintelligencetooureveryday environmentandmakeitsensitivetoourpresence.Itbuildsuponadvancesinsensorsandnetworks, pervasivecomputing,andartificialintelligence.Then,PersonalInformatics(PI)(Li,Dey,Forlizzi, 2010)isasetoftoolswhichuseincreasinglypopularwearabletechnologiesforacquiringpersonal informationonrelevantaspectsofpeople’sdailylives.Theyallowuserstoself-trackavarietyofdata aboutthemselves:fromuser’sphysicalstates(e.g.bloodglucoselevel),psychologicalstates(e.g.mood orstress),behaviour(e.g.movements)andhabits(e.g.foodintake,sleep)andcontextualinformation (e.g.peoplemet).Withself-trackingweintendsystematicrecordingandcollectionofpersonaldata, usuallybyusingvarioustechnologies,withtheaimtoimprovesomeaspectsofdailylife.Personal

Informaticssystemsbestshowanexpandingtrendinacquiringpersonaldataforprovidingsupport tousers,fromincreasingtheirself-awarenesstoimprovingtheirself-knowledgeandmotivatingself-reflection(Rapp&Cena,2014;Rapp&Tirassa,2017).AlthoughPIsystemsmayprovideuserswith compleximagesofthemselves,offeringopportunitiesforpersonalgrowthandchange,aswellas healthmanagement,self-trackingalsointroducestheproblemofinformationoverloadandtheneed tofilterthecollecteddata(Rappetal.,2016).TheinfluenceofthePIsystemscanbeoverlyinvasive anddominantinaperson’slife.Also,somesuggestionscouldarriveininappropriatemoments, posingsomeprivacyconcerns. QuantifiedSelf(QS)orlifeloggingisamovementwhichtriestointegratedevelopingtechnology intotheprocessofpersonaldataacquisitionandanalysisandreliesheavilyonIoTandPItools (Marcengo&Rapp,2013).Withsuchtechnologicaladvances,theamountandthetypeofdatathat canbecollectedaboutusersisexponentiallyincreasing,creatingaconstantstreamofinformation thatmayrevealmanyaspectsoftheirdailylives.ThisiswhenUserModeling(UM)processcometo playsinceitcreatesdigitalrepresentationsofusers,usedtocalibrateandadapttheinteractionwith asystem,ortoprovidepersonalizedrecommendations(Brusilovsky,Kobsa,&Nejdl,2007).Until recently,UMfieldhasfocusedmainlyonwebdata(Kobsa,Koenemann,&Wolfang,2001),thus creatingUserModelsusingdatafromwebusagebehaviourandprovidingprimarilypersonalizedweb applicationsandsystems.Asmanyreal-worddata(hoursofsleep,bloodpressurelevels,locations visited,etc.)arebecomingeasilyavailableduetoself-trackingmethodologiesandPItools,interesting opportunitiesforusermodelingarise.UM’scouldbeenrichedwithdataconcerningusers’real-worldcharacteristicsandactivitiesandcouldincorporatesuchinformationtocreatean“enhanced” modeloftheuser,i.e.aholisticrepresentationofherbody,activities,andhabits.The“enhanced” usermodelcouldenablehighlypersonalizedservices,sincethepersonalizationprocesswouldbe basedonrealhumanbeingsratherthanonWebbehaviour.Thiscouldfurthersupportnewforms ofpersonalizedandhighlydynamicservicesdirectlyintegratedintheusers’reallives,byadapting themselvesalmostinreal-timedependingontheongoingusers’internalstatesandexternalcontexts. So,theinevitablesymbiosisbetweenUserModelingandPIsystemsisbeingborn.Ononehand, UMcanbefilledwithdatacomingfromPIsystems,providingmoreeffectiveadaptiveservicesin UM-basedapplications(Cena,Likavec,&Rapp,2015).Ontheotherhand,PIsystems,asaresultof UMusage,canprovidepersonalizedvisualisationsandrecommendationsthatimprovetheirefficacy inpromotingbehaviouralchange. Inthisscenarionewresearchquestionsarise:WhatkindofpersonalizationenabledbyPIsystems wouldbethemostusefulintheusers’eyes?Whataretheopportunitiesforusermodelingcoming fromPIsystems?Whatworrieswouldtheusershaveregardingpersonalization?Hence,thereis aneedoffurtherresearchinthefieldofpersonalization.Toanswerthesequestions,weusedaPI system,developedpreviouslybyourresearchteam,togroundauserstudy.Thesystemcanbeused tocollectandaggregatedatafromdifferentself-trackingtoolsandprovideusefulvisualisationsand correlationstousers.Thefocusofthispaperisnotonthesystemitselfandonthedesignchoices madewhenbuildingit,ratherontheuserstudyenabledbythesystem.Thesystemwasusedto triggerreflectionsonthefutureofpersonalizationandonhowPIdatacouldenablenewformsof recommendation.Ourfinalaimwastogatherinsightsfromusersonhowtobuildan“enhanced”UM fordesigningnovelpersonalizedservices.Asopposedtootherworksinthisfield,ourstudywasnot conductedwiththeaimofdesigningaparticularsystem,rathertogatherknowledgewhichcouldbe usedbyawidespectrumofresearcherswhendesigningpersonalizedservicesandwhichwouldtake intoaccountusers’perspectiveonfutureofpersonalization. RELATEd woRK Boththecommercialandresearchcontextsarewitnessingrapidincreaseinthenumberofavailable mash-upsystemsandotherserviceswhichcanenableandsupporttheQuantifiedSelfprocess.These

servicesareusedtocollectandaggregateuserdataandpresentthembacktotheuser.Manyofthese services,suchasGoogleFit1_{(FitnessData),AppleHealth}2_{(Fitness/HealthData),Tictrac}3_{(Health}

Data),Headsuphealth4_{(Fitness/HealthData),Beeminder}5_{(Goals/ProductivityData)focusondata}

describingone’sphysicalstateandusuallyregardonlyasingledomain(health,fitness,etc.),with limitedaggregationfacilities.Thismeansthattheysimplyimportthedatafromdifferentsensors/ applicationsintoasingleuseraccount. Astepforwardistooffermeaningfulaggregationsandcorrelationstousers.Thisisdoneby someacademicprojects,asthefollowingones.Salud!(Medynskiy&Mynatt,2010)aggregatesdata fromdifferentsourcesbutletsusersinterpretfreelythegathereddata.Finally,Bentleyetal.(2013) designedahealthmashupsystemwhichemphasizessomerelationsbetweendiversebehavioural datausingnaturallanguage. However,noneofthesesystemsexploitsthebigamountofpersonaldatatoprovidepersonalized featuresorenablepersonalizedservices(suchasrecommendationsoradaptiveinformationprovision/ adaptiveinterfacereconfiguration).OneexceptionistheworkofSchmidt,Benchea,Eichin,and Meurisch(2015)whoproposedtosupportusergoalachievementthroughthecombinationofgoal datawithperformancedata,i.e.theyprovidepersonalizedtrainingplansbasedonthegathered performanceinformation.AnotherworkinthisdirectionistheworkofLee,Kim,Forlizzi,andKiesler (2015)whopromptedtheuserstoreflectontheirowngoalsencouragingthemtopersonalizetheir trainingprogramsbythemselves.WecanalsomentionGoogleNow,anintelligentpersonalassistant developedbyGoogle.Alongwithansweringuser-initiatedqueries,GoogleNowpro-activelydelivers theinformationthatitpredictsandthattheusersmightwantandneed,basedontheirsearchhabits anddataobtainedfromtheirsmartphones(location,interests,calendarentriesetc.).Thus,itprovides suggestionthatarecontext-awareanduser-dependent.Similarly,therearesomeactivity/foodtrackers (e.g.MyFitnessPal6_{,SuperTracker}7_,LoseIt8_{)whichmakerecommendationsaboutnutritionandgoals,}

buttheyprovideonlysimplestaticpersonalizationbasedonuser-compiledprofile,andnotdynamic recommendationscreatedontheflybasedoncontextanduserfeaturesdetectedinrealtime.W.r.t. emotionalcontent,wecancitetheworkofNielek&Wierzbicki(2010)whichpresentsaframework foremotionallyawaremobileapplication.Thedatacontainingtheinformationabouttheemotions (SMS’s,backgroundnoises,etc.)arecollectedandusedtodeterminetheuser’smood,basedonwhich theapplicationcanbehavedifferently,butnorecommendationsareprovidedtotheuser. Anothercloselyrelatedfieldislifelongusermodeling.Inlifelonglearnermodelingthedata abouttheusers’learningbehaviourisusedforthecreationandmanagementoflifelonglearner modelswhichcapturevariousaspectsabouttheuser’slearningactivityoverlongperiodsoftime. Kay(2008)providessuchaframeworkwheredataforlearnermodelingisobtainedfromwearable sensors,interactivelearningactivities,Webtraceanalysisandexplicitinformationprovidedbythe user.Thesedataarescatteredacrossmanydevicesandstoragemediaandusedtoreasonaboutthe valuestocreateacompleteUserModel.Elliottetal.(2009)provideageneralarchitecturewhichcan gather,storeandprocessdatafromvariousdatasources.Theiraimistocapturetheheterogeneous dataregardingvariousaspectsofaperson’slifeandextracttherelationsbetweendifferentfeatures whichwouldimproverecommendationofrelevantinformation.ARBUT(Hohwald,Frias-Martinez &Oliver,2010)isasystemwhichcanprocesslargeamountsofheterogeneoussensordataand producenumerouscomplexUserModel.Someofitsreusablecomponentscanbesharedinvarious applicationdomains.Finally,PersonisAD(Assadetal.,2007),isaframeworkwhichmodelspeople, sensors,devicesandplacesandprovidesthisinformationtootherapplications.Theuserscancontrol theinformationbeingmodelledandhowitisbeingused. AlthoughmanyuserstudieshavebeenconductedtoinvestigatehowPIsystemsimpactonusers’ everydaylives,noneofthemhasspecificallyinvestigatedhowpersonaldatacollectedbyPIsystems couldimprovepersonalization.Asexamplesofuserstudiesofthistype,wecanciteLietal.(2010), whostudiedhowexperiencedusersusePIsystems,Choeetal.(2014),whoinvestigatedQuantified Selfers’practices,focusingonpeoplemotivatedintrackingpersonaldatainordertoseewhateffects

ontheirbehaviour;Fritzetal(2014),whostudiedinthefieldthelongtermusageofwearabledevices forphysicalwellness,focusingonthebehaviourchange;Rooksbyetal.(2014)whoinvestigatedthe trackingprocess,identifyingsomedifferent“personaltrackingstyles”;Rapp&Cena(2016)who focusedonhowpeoplewithoutpreviousexperienceinself-trackingusePIsystems.Finally,we mentionsomefewworkspartiallydevotedtostudypersonaldataforpersonalization.InBarua,Kay &Paris(2013b),authorsinvestigatepeople’sperceptionofcontrolovertheirdata(scrutability).In Barua,Kay&Paris(2013a),automaticdatacaptureoverlongperiodsoftimeandcontroloverdata storageareexploredfromtheusers’perspective.Moreover,Warshawetal.(2015)studiedhowa systemthatletsusersseetheirpersonalityisperceivedbythem.Participantsfoundtheirpersonal traitsprofilescreepilyaccurateanddidnotlikesharingtheminmanysituations.However,theywere enthusiasticaboutthepossibilityofbeingprovidedwithpersonality-basedrecommendations.Such studies,nevertheless,werenotconnectedtoPIfield. USER STUdy Inthissection,wereporttheresultsofouruserstudyinwhichourmainfocuswasonobtaining insightsonhowPIdatacouldbeexploitedinthefuturetoprovidenewtypesofpersonalizedservices andrecommendations.Tothisaimweinvitedtheparticipantstocollectpersonaldatathroughself-trackingtoolsandvisualizetheminasystemdevelopedpreviouslybyourresearchteam,inorderto anchortheirinsightsinarealusecase.AsmentionedintheIntroduction,thefocusofthispaperis notonthedesignofthesystemitself,ratherthesystemwasusedtotriggerreflectionsonthefuture ofpersonalizationandonhowPIdatacouldenablenewformsofrecommendation. System Thesystemweusedinourstudywasdevelopedpreviouslybyourresearchteam.Herewereport brieflyitsmainfunctionalities.Thesystemisableto: • Gatherandintegrateheterogeneoustypesofpersonaldatafromvariouswearabledevicesand PImobileapps;itaggregatesdifferentdatachannels,suchassleepandstepsfromJawboneUp braceletorWithingsActivitésmartwatch,skinarousal,heartbeatandbodytemperaturefrom EmpaticaE3bracelet(https://www.empatica.com/),aswellaslocations,transportmeans,calories andfoodintakesfrommobileappslikeMovesandMyFitnessPal; • Findsimilaritiesandcorrelationsamongdata; • Visualizethegathereddata. Inparticular,thesystemprovidestwodifferentkindsofvisualization.Inthefirstonetheuseris providedwithasnapshotofaspecifictrackedparameterusingcolourcodes.Bypositioningherself infrontofthescreen,theusercanseeadata-drivenimageofherself,consistingofthedataofa specificchannel(e.g.herheartratetoday).Inthesecondvisualisationtype,thesystemprovidesan overviewofallthedifferentdatacollected,displayedsimultaneouslyonparalleltimelines.Theuser canalsovisualizethelocationinwhichshewasinaspecificmoment,andthephotosshemadeina specificplace.Suchinformationisaimedtocontextualizethedatasheislookingat. However,acompleteillustrationofthesystem’sfeaturesisbeyondthescopeofthispaper. MoredetailsaboutthesystemcanbefoundinRappetal.(2017).Inthiswork,ithasbeenusedonly togeneratereflectionsonthefutureofpersonalization.Theaimofthisworkistoreflectonhow personalinformaticssystemscouldenablenewpersonalizedservices,andnottodescribeorassessthe system’sfunctionalities.Thisresearchfollowsapreviousstudywhereweevaluatedthesystemwith referencetousabilityandcompliancetoaseriesofuserrequirements:inthatoccasion,weincidentally gatheredsomeinitialinsightsalsoonpersonalization(Rappetal.,2017).Thismotivatedthissecond

studywhichdealswithcompletelydifferentresearchquestionsandprovidesnewfindings:here,we arenotinterestedinevaluatingoursystem,butinusingitasatriggertothoroughlyinvestigatehow personalizedservicescouldevolveinthefutureduetotheavailabilityofPIinstruments. Method Weconductedaqualitativestudyusingathinkingaloudprotocol,semi-structuredinterviewsandfocus groupstoinvestigatehowusersmightperceiveandenvisionthefutureofpersonalizationtriggeredby theavailabilityofdifferentpersonaldata.Thinkingaloudisauserstudyprotocolwhereparticipants havetoverbalizeeverythingrelatedtoagivensystemcomingtotheirmindduringitsusage.Although thismethodcanprovideinsightfulqualitativeresultsitmayrisktodistracttheuser,jeopardizingthe gatheringofdatarelatedtotheefficiencyandefficacyofthesystemunderevaluation.Wechosethis methodologysincewewerenotinterestedinassessingourPIsystem’sefficacyandefficiency:instead, wewantedtousesuchsystemtogenerateinsightsaboutthefutureofpersonalizationanchoredina realcaseofuse.Therefore,atechniquethatleavesparticipantsfreetoexpresstheirthoughts,needs anddesireswhilecompletingataskwasconsideredthemostappropriate.Semi-structuredinterviews wereusedtogathermorefocuseddata.Finally,focusgroupswereusedtogeneratefurtherinsights fromgroupdiscussion. Sample Ourstudyinvolved18participants(ageM=29,5,SD=8,6,females=10).Thesamplesizewas setaccordingtothestandardsinqualitativeresearch(Marshall,1996;Marshall,Cardon,Poddar,& Fontenot,2013).Nineparticipants(P1-P9)didnothaveanypreviousexperiencewithPI,sincewe wantedtoinvestigatealsohowindividualsthatarenotusedtocollectdataaboutthemselvesperceive thenewopportunitiesofferedbyPIdeviceswithrespecttopersonalization.Theremainingparticipants (P10-P18)havebeenusingself-trackinginstrumentsforalmostsixmonths. Procedure EachparticipanthadtouseawearabledeviceandasetofPIappsforfourweeks.Wegavethe participantsthepossibilitytochooseamongthreedifferentPItoolsets.Wewantedtorepresentdifferent casesbyprovidinguserswithdifferentself-trackinginstruments,accordinglytotheirinterest.This increasestheheterogeneityofthesample,allowingthegeneralizationofourresults(Gobo,2008). However,atthesametimeweneededtopartiallyhomogenizethesampleforthestudypurposes: leavingtheuserscompletelyfreetoselectthePIinstrumentswouldhaveexcessivelymultiplied theheterogeneityofthecasesunderstudy.Wefinallysetthreegroupsoftools,allowingustohave groupsofsixparticipantseach. Duringthestudy,wemettheparticipantstwotimes.Duringthefirstmeeting,wesetthewearable devicesandtheapplicationsontheirsmartphones,andprovidedinstructionsforcollectingdata.After fourweeks,theycametoourlab,transferredthegathereddataintooursystem,andwereprovided withthevisualizations.Duringthistime,theycouldfreelycommentonoursysteminathinking aloudsession.Then,afterthetrial,weinterviewedparticipantssingularlyfor30minutes,askingthem toreporttheirfirstimpressionsaboutthepossibilitiesthatallthesedatacouldopeninthefuture. Sincetheinterviewsweresemi-structured,questionswerethesameforalltheparticipants,butthey wereleftfreetodivertfromthetrackandexplorethethemesthatwehadnotforeseeninadvance. Weaskedtheparticipantsthefollowingquestions:Whatkindofdatayouconsidermostusefulto enablenewpersonalizedservices?Whatkindofopportunitiesandthreatsdoyouseeintheevolution ofpersonalizationtriggeredbyself-trackinginstruments?Whatkindofpersonalizedservicesyou considerthemostsuitabletoyoueverydayneeds? Finally,wedividedparticipantsintothreegroups(P1-P6,P7-P12andP13-P18)invitingthem tocollectivelydiscussopportunitiesforfuturepersonalizedservices,enabledbythedatacollected bywearableandmobiletechnologies.Participantswereallocatedtoeachgroupaccordinglytotheir

experienceinusingself-trackingtools:onegroupofexperiencedusers,onegroupofinexperienced users,andonegroupwherehalfoftheparticipantshadpreviousexperiencewithPItools,andhalf oftheparticipantshadnopreviousexperiencewithPItools. Groupdiscussionslastedforabout90minutes.Theresearcheronlymanagedtheturn-taking withoutinterveninginthediscussion.Astheparticipantshadtheopportunitytogathertheirowndata andputthemintouseinoursystem,thediscussionsweregroundedonarealusecase.Interviews andgroupdiscussionswereaudiorecordedthroughaportableaudiorecorder. data Collection P1-P3andP10-12usedanEmpaticaE3braceletfortrackingarousal,bodytemperature,andheart rate,MovesappforkeepingtrackoflocationsandSleepBotforsleeptracking;P4-P6andP13-P15 weregivenWithingsActivitésmartwatchfortrackingsleepandsteps,Movesfortrackinglocations andExperealformoodtracking;P7-P9andP15-P18weregivenMisfitShinebracelettotracksleep andfitnessactivity,Movesfortrackinglocationsandanappdevelopedadhoctokeeptrackofthe phonecalls,emailsandSMS’ssent.Weselectedthewearabledevicesandthemobileappsaftera deskanalysis.Wechoseaspecificappordevicebecauseweconsidereditthebesttoolfortrackinga specificparameter.Forexample,Experealwasjudgedthebestappfortrackingmood,sinceitprovides anintuitiveinterfaceand,atthesametime,areliablewaytospecifythemoodbyusingascalefrom1 to10.Otherappswereconsiderednotusable,requiring,forexample,tocollecttoomanyparameters forassessingthemood,thusrequiringburdensomeactivityontheuser’spart.Participantshadto keeptheappsactiveandwearthedevicesforthewholeperiodofthestudy.Specifically,thismeant tochargethedeviceeverydayandwearitduringdayandnight.Appsweresupposedtobeusedat leastonceaday,reportingandvisualizingthecollecteddata.Participantsreportedhighcompliance infollowingthestudy’sinstructions.Only3participants,duringtheinterviews,reportedtohave forgottentowearthedeviceforoneentiredayormore. data Analysis Weanalysedthedatacomingfromthethinkingaloudsession,theinterviewsandthegroupdiscussions togetherusingathematicanalysis.Dataanalysisfollowedopenandaxialcodingtechniques(Strauss andCorbin1990).Resultswerecodedindependentlybythefirstandthesecondauthor:theywere brokendownbytakingapartsentencesandbylabellingthemwithacode.Then,wecomparedthe outcomesdiscussingandmanagingalltheinconsistencies(MacQueen,McLellan-Lemal,Bartholow, &Milstein,2008).Userswerenotgivencompensationfortheirparticipation(seeTable1). RESULTS Havingalltheirdataavailableinasinglecentralizedlocation,participantsperceivedthesystemas beingcapableofknowingdifferentaspectsoftheirdailylivesand,onthebasisofthisknowledge, wishedforaseriesofservicestailoredtotheiridiosyncratichabits,preferences,activities.Results

Table 1. The user study process

1 Step 2 Step 3 Step 4 Step

-Wearablesandappsare

providedtoparticipants -Participantsareprovidedwithvisualizationoftheir datawithoursystems

-Participantsare

interviewedindividually -Commentsandinterviewsareanalysed andcodedbyauthors -Participantsusedevices

andappsfor4weeks -Participantscancommentfreelythroughathinking aloudsession

-Participantsaredividedin 3groupsanddiscussabout personalization

showthatparticipantswishedforrecommendationsthatcouldconsiderdifferentaspectsoftheirlife, from“internaldata”,likecognitivestatesandemotions,tocontextualfactors.Moreover,participants highlightedthepotentialrisksthatthepervasivenessofdatacollectioncouldentailforthem.We nowreportourfindingsbydividingthemintothemajorthemesemergedduringtheanalysis.While Table2mapstheresulttotheinitialresearchquestions.

Personalized Goals Setting

Participantsalreadyusedtoself-trackingfocusedtheirconsiderationsandexpectationsmainlyon wellbeingdomain.Theyemphasizedthatthesystemshouldhelppeoplesetpersonalizedgoalsrelated totheirphysicalactivityandlifestyle,withtheaimofimprovingtheiroverallhealth.P11,forexample, emphasizedduringtheinterview:“I’dliketohavepersonalizedfutureobjectives…Idon’tknow howitcouldbedone,butIexpectthatitcouldforeseewhatwouldbeusefulformeinthefuture onthebasisofmyactualcondition,suggestingalsohowtoreachmygoals”.Themajorityofthese participants(5outof9)highlightedalsothatitwouldbedesirableifpersonalizedsystemswould workwiththemtodefinegoalsandplans,asiftheyweresomesortofcompanions,ratherthanmere top-downprompters.Theseparticipantsagreedonthispointduringthegroupdiscussion,confirming thatthepossibilityofachievingspecificobjectiveswiththehelpofgoal-basedrecommendations couldrepresentanimportantstepforwardinpersonalizedservices.

Holistic data Collection

Severalparticipants(8outof18)alsoemphasizedthatitwouldbepromisingtowidenthetypeof informationcollectedbythesystemforcreatingacompletedigital“double”,aholisticrepresentation oftheuser,onthebasisofwhichtobuildpersonalizedservices.Theideaofexpandingthekindof trackeddataandmovingbeyondwellbeing,camemainlyfromthosewhodidnothaveanyprevious experiencewithPItools(6outof9)duringthegroupdiscussion.Theseparticipantswerenotfocused (astheremainingones)onbehaviourchange,andexploreddifferentcontextsinwhichpersonalization couldimprovetheirdailyactivities.Theywerestruckbytheopportunitytotrack“internalparameters” suchasphysiologicalstatesandemotionallevels,especiallywhenthiswasdoneautomatically. Physiologicalparameterswereconsideredusefulformedicalpurposes,butsomeparticipants,like P3,perceivedthemasanaidtoimprovedailynutrition,forexamplebysuggestinghowmanycoffees sheshoulddrinkdependingonherarousallevel.Othersreflectedonhowcontinuousself-tracking ofemotionalstateswouldenablerecommendationsbasedalsoontheirunconsciouspreferences andwishes,supportingbettertheirdecision-makingprocessandtheirchoicesduringtheireveryday activities.P2,forexample,suggestedthatbyusingemotionaldatathesystemcouldmakesuggestions withoutrelyingonherrationalthinking,butexploitinghervisceraltastes,whichsheconsideredmore sincere.Thiscouldhappen,forexample,beforebuyingsomethinginastore,assuggestedbyP3,or whenchoosingabooktoread,orafriendtocall,assuggestedbyP6andP7.Alltheseparticipants perceivedthiskindofemotion-basedrecommendationsclosertowhattheyreallywantandfeel,thus

Table 2. Answers to research questions

Research Questions Answers from the Study

WhatkindofpersonalizationinPI systemswouldbethemostusefulin theusers’eyes? -Personalizedgoalsetting -Contextualizedrecommendations -Dataexchangeandsharingforpersonalization“ondemand” Whataretheopportunitiesforuser modelingcomingfromPIsystems? -Holisticdatacollection Whatworrieswouldtheusershave

theyweremoreinclinedtofollowthem.Ontheotherside,otherusersimaginedhowhavingcognitive data,suchaslevelofattention,interest,mentalworkload,couldproviderecommendationsonthework taskstobeprioritized,orwhatandhowtheyshouldstudy,improvingworkingandlearningactivities. Contextualized Recommendations Ifphysiological,emotionalandcognitivedatamayservefordifferentkindsofpersonalizedservices, participants(13outof18)expectedthatafuturepersonalizedsystemcouldcombinethemalltotake intoaccountalltheconditionsthatmayaffectaparticularactivity.“Whensuggestingtometolengthen myrun,thesystemshouldbeawareofmycurrentmood,myhealthcondition,forexamplemyactual heartbeat,andformulateasuggestiononthebasisofthisinformation…forexampletellingmethat Ishouldrunfortwentyminutesmoretoreachmygoal,butwithalowerpace,becausemyheartrate istoohigh,andfollowinganotherroute,becausethisstreetismakingmenervous.”,saidP14during thegroupdiscussion.“AndmaybealsothatIshouldlowermymusicbecausethetrafficisgrowing andmylevelofattentionislowering.”,addedP17inthesamegroupdiscussion.Thispointstothe needofconsideringthewidercontextinwhichtherecommendationisprovided. Mostoftheparticipants(14outof18)believedthatduetotheavailabilityofallthese personaldataandtheirintegrationinauniquesystem,personalizationcouldbecomepervasive andintegratedintheireverydayactivities.Someofthempointedoutthepossibilityofsuggesting alternatives to their current behavior based on the system’s knowledge of their habits, the contextualconditionsandtheconsiderationsofwhattheywilldointheimmediatefuture.P9,for example,duringtheinterview,said:“Itshouldbepossibletorecommendtomenottotakethe cartoday,asIamusedto,buttousethebicycle,sincethisnightIneedtogotobedearlygiven mytravellingtomorrow,anditwouldbebettertobeabitmoretired.”.Others,likeP15,when interviewedstressedtheneedofhavingthesesuggestionsjustbeforemakingadecision:“Ifthe systemknowsthatIgoouteverydayat6pmforawalkandatthesametimerecommendstome towalkmore,itshouldsuggestitalongwithanotherroutetofollowformydailywalkjustin time,namelywhenI’malreadyoutandabouttodomydailyactivity”.Alltheserecommendations wereexpectedtobedeliveredforthetargetactivityattherighttimeandattherightplaceby almostalltheparticipants(16outof18).“ThegreatpotentialthatIseehereisthatthissystem canalwaysknowwhatI’mdoing.Soitshouldsuggesttomewhattodowhenit’sreallyneeded”, saidP3duringthegroupdiscussion.

data Sharing and Exchange

Participants(7outof18)highlightedthatitwouldbeusefultomakeallthedatacomingfrom PItoolsexchangeablewithotherapplications,withthesharingofthedataundertheircomplete control.“Itshouldbepossibletouseanexistingserviceorapplicationandmaybepourallthe knowledgerelatedtomeintheretogetsomethingthatistailoredonlytome.Idon’tknow… likeprovidingallthedatarelatedtomymobilitybehaviorwhenbuyingacarandhaving recommendedtherightcarforme.ButonlywhenIwantitandforthatparticularmomentor need.”,notedP14duringtheinterview.Thisattitudeshowsthatuserswishedforsharingtheir dataamongdifferentplatformsandservices.However,duringthegroupdiscussionitemerged thattheywantedtobetheownersoftheirownknowledgeaboutthemselves,aknowledgeto injectinexistingsystemsforhavingpersonalizedoutcomesforspecificandmomentarygoals, inasortofpersonalization“ondemand”.Infact,themajorityofparticipants(16outof18) stressedhowthecontroloftheirdatashouldremainintheirhands,confirmingascepticism, ormorepreciselyaworry,inthewidedisseminationofpersonalinformation.Forthisreason, theywantedtostorethedataontheirlocaldeviceratherthanonaremoteserver.Exceptions wererepresentedbytwoparticipantswithpreviousexperienceinself-tracking,whodidnot haveanyprivacyconcerns.

Control Anothersharedconcern(11outof18)wasrelatedtothecontrolexertedbythesystemoverthe informationmadeavailable,aswellasthereductionofchoice.P16,forexample,duringthegroup discussionemphasizedthat“Suchrecommendationsareusefulwithoutanydoubts.Butthesystem mayalsorecommendwhatisusefulforit…Therecouldbemanycommercial,orpolitical,reasonsfor whichitcouldrecommendmesomething”.WhileP18addedthat“Itisasifalltheserecommendations designacirclearoundmethatdefineswhatisrightforme…whatisoutsidethecircleiscompletely subtractedfrommysight.Thiscouldbeaproblem…notonlybecauseofasortofcontrolthesystem hasoverus,butalsobecauseoftherecommendationsthemselves…Imean,sometimesmaybeIlike thingsthatIdonotexpectatall,orthingsthatarenotaddressedtomywellbeingortosatisfymein thehabitualway…howcouldthesystemdealwithsuchdesires?”.

dISCUSSIoN ANd dESIGN IMPLICATIoNS

Ourstudyshowsthatpeopleexpectadvancedformsofpersonalizationfromsystemsthathavecertain informationaboutthem.Thisseemstobearequirementforallthetoolsthatgatherpersonaldata, andconsequentlyforallPIsystems.MostoftheuserswithoutpreviousexperiencewithPItools(8 outof9)wouldhavecontinuedtousetheself-trackingtoolsifsuchinstrumentswouldbeableto offerthemsomeusefulrecommendations.Thissuggeststhatcontentpersonalizationcouldrepresent akeyvalueforuserswhoarenot“behavior-changeoriented”,andactuallydonotfindsufficient motivationtobeengagedinself-tracking(Perera,Zaslavsky,Christen,&Georgakopoulos,2014). However,experiencedself-trackersprovedtobefocusedmoreongoals,confirmingthepreliminary findingsfromourpreviousstudy,whereexperiencedparticipantshighlightedtheneedofhaving personalizedplanstoachievebehaviourchangegoals,especiallyinthelongterm(Rappetal.,2017). PersonalizedtechnologiesbasedonPIdatahavealsobeenimaginedascapableofsatisfying users’innerdesires:thepossibilityofreceivingrecommendationsbasedontheiremotionsand ontheirunconscious,non-rationalintentionswerewelcomedwithenthusiasm,asawayforbeing moresatisfied.ThissomehowcontraststhefindingsofWarshawetal.(2015)whoemphasizedhow being“known”intheirinnertraitsbyanartificialentityisperceivedas“creepy”byusers.Thisis likelyduetothefactthatourstudyfocusedonpersonalizationmatterswhiletheWarshawetal.’s studyfocusedonusers’profiles.Being“scanned”byacomputersystemmaybeperceivedmore favourablyifthefocusisonthepersonalizedservicesitmayenable,asalsotheWarshawetal.’s studynoted,byhighlightingthattheiruserswereoverwhelminglypositiveaboutgettingactivityor eventrecommendations. However,benefitsofpersonalizationwerenotconsideredrisk-freeevenbyourparticipants.The participantsespeciallyemphasizedthefearofahiddenpersuasion.Evenaformofsoft-paternalism (Thaler&Sustein,2008),inwhichrecommendationscouldbeusedtosubtlytransmitwaystobehave, notwelcomedotherwise,raisedmanyconcerns.Thisriskhasbeenalsonotedinbehaviourchange technologies,whichmayprovidegoalsnotchosenbytheindividual,ratherreproducingvaluestaken forgrantedinagivensociety(Purpuraetal.,2011).Forthisreason,personalizedtechnologieshave beenperceivedmorefavourablyifactingascompanions,committedtoworkwiththeusers,inorderto findrelevantgoalsandplansforthem.Moreover,participantsalsohighlightedthatrecommendations couldincludeonlyasmallsubsetofthemultiplepossibilitiesforactingandknowingthattheusers mayhavebyaccessingalltheirpersonaldata,hidingsomehowalltherest.Onthebasisofthese resultsweoutlinedaseriesofsuggestionsfordesigningnovelPIsystems,abletoprovidepersonalized servicesthatcanmeetusers’expectations: • CreateacompleteUMonthebasisofthePIdatacollected.Almostallinexperiencedparticipants andaminorpartofexperiencedoneswishedforreceivingrecommendationsbasednotonly onspecific,singularparameters,butalsoonaholisticrepresentationofthemselves.Different

researchinUM,triedtouseemotional(Tkalčič,Burnik,Odić,Košir,Tasič,2012),environmental (Adomavicius&Jannach,2014),andphysiological(Bennett&Quigley,2011)datatoprovide tailoredrecommendations.However,noattemptsweremadetocombinesuchdatainorderto buildacomplete,dynamic,andsituated(inthecurrentcontext)mirroroftheuserthroughwhich givingjust-in-timerecommendationsispossible.UMresearcherscouldexploretechniques employedincontext-awaresystemsinpervasiveenvironment(Perera,Zaslavsky,Christen,& Georgakopoulos,2014),todealwithdataheterogeneity; • Supportreasoningonthecollecteddata,tosuggestalternativesbasedontheuser’sactualcontext. Themajorityofparticipantsexpressedtheneedofhavingrecommendationsonthebasisof theircurrentsituation,eithertheenvironmentalcontextortheir“internalstates”.Thiswould enable,intheparticipants’eyes,thepossibilityofbeingprovidedwithalternativestotheirusual behaviouralpattern.Newpersonalizedsystemsshouldformalizetheknowledgetheycollectto enablereasoningoverthedataandsuggestdifferentalternativesthatmayfitdiversesituations. Ontologies,forinstance,mightmaptheconceptsinaspecificdomainenablingontological reasoningwhichcouldexpandtherangeoftherecommendationsprovided(Cena,Likavec, Rapp,&Marcengo,2016).Wecouldsuggestabehaviourthatissimilarordifferentbutsomehow relatedtowhattheuserisusedtodoinginordertomaximizethebenefitofthesuggestionsgiven. Mobilesensingcapabilitiescouldfurtherhelpinrecognizingtheuser’scurrentcontextandadapt therecommendationsaccordingly.Forexample,wemayknowthattheuserlovesrunning,but accordingtoherhistoricaldatathisactivityiscorrelatedwithbadsleep.Sincetodaysheneeds tosleepwell,asshehasplannedalongtripforthenextday,thesystemmightsuggestasimilar activity,suchasswimming,tosatisfyherneedofdoingsport.Thissuggestioncouldbeprovided inproximityoftheswimmingpoolbyrecognizingtheuser’scurrentposition; • UsetheUMtohelpsetobjectivesandwaystoachievethem.Allexperiencedparticipantsfocused onthepossibilityofhavingpersonalizedgoalsandplans,tendingtouseself-trackingdevicesfor achievingspecificgoalsthattheyhaveinmind.Tosupportthiskindofusersthesystemcould setlong-termgoalsbasedontheuser’spastbehaviourandonthepredictionoffuturesituations. Itcouldproviderecommendationstriggeredbytheuser’scurrentcondition,suggestingwhich kindsofactionsandchangestheusershouldputinplacetomeetthesetgoals.Suchplansshould bebuilttogetherwiththeusers,asthemajorityoftheexperiencedself-trackersnoted,ina continuousexchangeofperspectivesbetweentheuserandthemachine,asaformofcooperative actionorchestrationwheresometimesitistheusertodrivethecooperation,othertimesitisthe machine(Ohlin&Olsson,2015); • Makethestoreddata“exchangeable”amongdifferentservicesandapplicationsbutonly forspecificpurposes.Participantsexpressedtheneedofexchangingdataamongdifferent applications.Usersshouldbeabletousealltheirpersonaldata(orasubsetofthemrelatedto aspecificparametertracked)inothersystems.This“datainteroperability”(Carmagnola,Cena, &Gena,2011)wouldsupplydigitalapplicationsandserviceswithadditionaldataaboutthe users,gatheredintimeduetothecapabilitiesofmobileandwearabletechnologies,aswell asstoredintheUM,enablingsupplementarypersonalizedfeatures.Thissortof“portable personalization” could allow users to have adaptive interaction modalities and tailored recommendationspracticallyeverywhereatanytime,apartfromthespecificsystemorservice theyareusing.However,asthemajorityofourparticipantsemphasized,usersshouldalways decidewhichkindofdatatheywanttoprovide,forwhatpurposesandforhowlong,inasort ofagoal-drivenandcontrolledpersonalization; • Allowfornovelty,serendipityandapossibilitytoexitfromthe“magiccircle”.Serendipityisa well-knownissueincontent-basedrecommendersystems,whichtendtoproducerecommendations withalimiteddegreeofnoveltyduetoover-specialization(Lops,deGemmis,&Semeraro,2010). PersonalizationbasedonPIdatashouldallowuserstolookoutsidethecirclecircumscribedby therecommendationsprovided.Ourparticipants,infact,expressedaseriousconcernrelatedto

thecontrolthatrecommendersystemcouldexertontheinformationprovided,actuallylimiting theinformationavailable.Thismeans,ontheoneside,thattheyshouldbeprovidedalsowith suggestionsnotdirectlytiedtotheirUM;ontheotherside,thattheyshouldhavethepermission toaccessthewholeloadofknowledgethattheirdataentail.Thiscouldbeachievedbyboth makingtheirUMscrutable(Kay&Kummerfeld,2013),andbygivingthemthemeanstoaccess theinformationandtheopportunitiesexcludedbytherecommender. CoNCLUSIoN Inthispaper,weemployedaPIsystemtoconductauserstudywiththeaimtounderstandusers’needs andexpectationsw.r.t.personalizationandrecommendation,particularlyenabledbythepresence ofmobileandwearabletechnologies.Weoutlinedhowusersthathavepreviousexperienceinself-tracking,aswellasindividualsthatarenotusedtotrack,envisionthefutureofpersonalization. Whileexperttrackersprefertohavepersonalizedplanstoachievetheirgoals,inexperiencedones wishforadigitaldouble,capableofaccountingforallthedifferentaspectsthatcharacterizetheir “self”,fromcognitivetoemotionalstates.Moreover,participantshighlightedtheneedforproviding recommendationsonthebasisoftheentirecontextthatcharacterizesanindividualinagivensituation. Allthesedata,intheparticipants’eyes,shouldbemadeexchangeableamongdifferentsystems allowingfor“pervasive”personalizedservices.However,participantsalsohighlightedprivacyand controlconcerns,whichcouldjeopardizetheacceptanceofsuchservicesinthefuture.Onthebasisof suchresultswepointedtofivedesignrecommendationswhichaimtoprovidesuggestionsonhowto satisfytheusers’needs.Inouropinionbothpractitionersandresearcherscouldusethesesuggestions asaguidefordesigningnovelpersonalizedserviceswhichwouldembedthemostwishedfeatures, aswellasavoidthemostpressingconcerns,pointedbydifferentcategoriesofusers.

REFERENCES

Adomavicius,G.,&Jannach,D.(2014).Prefacetothespecialissueoncontext-awarerecommendersystems.

User Modeling and User-Adapted Interaction,24(1),1–5.doi:10.1007/s11257-013-9139-2

Assad, M., Carmichael, D. J., Kay, J., & Kummerfeld, B. (2007, May). PersonisAD: Distributed, active,scrutablemodelframeworkforcontext-awareservices.InProceedings of the 5th International

Conference on Pervasive Computing (PERVASIVE ’07),LNCS(Vol.4480,pp.55–72).Berlin:Springer.

doi:10.1007/978-3-540-72037-9_4

Barua,D.,Kay,J.,&Paris,C.(2013a,April).Viewingandcontrollingpersonalsensordata:Whatdouserswant? InS.Berkovsky&J.Freyne(Eds.),Proceedings of the 8th International Conference on Persuasive Technology

- PERSUASIVE 2013,LNCS(Vol.7822,pp.15–26).Berlin:Springer.

Barua,D.,Kay,J.,&Paris,C.(2013b,November).Foundationsforinfrastructureandinterfacestosupport usercontrolinlong-termusermodelling.InProceedings of the 25th Australian Computer-Human Interaction

Conference: Augmentation, Application, Innovation, Collaboration (OzCHI ‘13)(pp.125-134).NewYork,NY:

ACM.doi:10.1145/2541016.2541034

Bennett,M.,&Quigley,A.(2011,July).CreatingPersonalizedDigitalHumanModelsofPerceptionforVisual Analytics.InProceedings of the 19th international conference on User modeling, adaption, and personalization

(UMAP’11)(pp.25-37).Heidelberg:Springer-Verlag.doi:10.1007/978-3-642-22362-4_3

BentleyF.,TollmarK.,StephensonP.,LevyL.,JonesB.,RobertsonS.,PriceE.,CatramboneR.,&Wilson J.(2013).Healthmashups:Presentingstatisticalpatternsbetweenwellbeingdataandcontextinnatural language to promote behavior change. ACM Transactions on Computer-Human Interaction (TOCHI), 20(5),30:1-30:27.

Brusilovsky,P.,Kobsa,A.,&Nejdl,W.(Eds.).(2007).The Adaptive Web, Methods and Strategies of Web

Personalization.Berlin:Springer.

Carberry,S.,&deRosis,F.(2008).IntroductiontospecialIssueonAffectivemodelingandadaptation.User

Modeling and User-Adapted Interaction,18(1),1–9.doi:10.1007/s11257-007-9044-7

Carmagnola,F.,Cena,F.,&Gena.C.(2011).UserModelInteroperabily:Asurvey.User Modeling and

User-Adapted Interaction,21(3-2),285-331.

Cena,F.,Likavec,S.,&Rapp,A.(2015,September).QuantifiedSelfandModelingofHumanCognition.In

Adjunct Proceedings of the 2015 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2015 ACM International Symposium on Wearable Computers (UbiComp/ISWC’15 Adjunct)(pp.1021-1026).NewYork,NY:ACM.doi:10.1145/2800835.2800954

Cena, F., Likavec, S., Rapp, A., & Marcengo, A. (2016, September). An ontology for quantified self: capturingtheconceptsbehindthenumbers.InProceedings of the 2016 ACM International Joint Conference

on Pervasive and Ubiquitous Computing: Adjunct (UbiComp ‘16)(pp.602-604).NewYork,NY:ACM.

doi:10.1145/2968219.2968329

Choe, E. K., Lee, N. B., Lee, B., Pratt, W., & Kientz, J. A. (2014, April). Understanding quantified-selfers’practicesincollectingandexploringpersonaldata.InProceedings of the SIGCHI Conference

on Human Factors in Computing Systems (CHI ‘14) (pp. 1143-1152). New York, NY: ACM.

doi:10.1145/2556288.2557372

Elliott,D.,Hopfgartner,F.,Leelanupab,T.,Moshfeghi,Y.,&Jose,J.(2009,June).Anarchitectureforlife-long usermodelling.Presented atWorkshop on the Lifelong User Modeling, UMAP 2009.

Fogg,B.J.(2003).Persuasive technology. Using computers to change what we think and do.SanFrancisco, CA:MorganKaufmann.

Fritz,T.,Huang,E.M.,Murphy,G.C.,&Zimmermann,T.(2014,April).Persuasivetechnologyinthe realworld:astudyoflong-termuseofactivitysensingdevicesforfitness.InProceedings of the SIGCHI

Conference on Human Factors in Computing Systems (CHI ‘14)(pp.487-496).NewYork,NY:ACM.

Gubbi,J.,Buyya,R.,Marusic,S.,&Palaniswami,M.(2013).InternetofThings(IoT):Avision,architectural elements,andfuturedirections.Future Generation Computer Systems,29(7),1645–1660.doi:10.1016/j. future.2013.01.010

Hohwald,H.,Frias-Martinez,E.,&Oliver,N.(2010,June).ARBUD:Areusablearchitectureforbuildinguser modelsfrommassivedatasets.Presented at theWorkshop on Pervasive User Modeling and Personalization

(PUMP), UMAP ‘10. Kay,J.(2008).Lifelonglearnermodelingforlifelongpersonalizedpervasivelearning.IEEE Transactions on Learning Technologies,1(4),215–228.doi:10.1109/TLT.2009.9 Kay,J.,&Kummerfeld,B.(2013).Creatingpersonalizedsystemsthatpeoplecanscrutinizeandcontrol:Drivers, principlesandexperience.ACMTrans.Interact.Intell.Syst.,2(4). Kobsa,A.,Koenemann,J.,&Wolfang,P.(2001).Personalizedhypermediapresentationtechniquesfor improvingonlinecustomerrelationship.The Knowledge Engineering Review,16(2),111–155.doi:10.1017/ S0269888901000108

Lazar,A.,Koehler,C.,Tanenbaum,J.,&Nguyen,D.H.(2015,September).Whyweuseandabandonsmart devices.InProceedings of the 2015 ACM International Joint Conference on Pervasive and Ubiquitous Computing

(UbiComp ‘15)(pp.635-646).NewYork,NY:ACM.doi:10.1145/2750858.2804288

Lee,M.K.,Kim,J.,Forlizzi,J.,&Kiesler,S.(2015,September).Personalizationrevisited:areflectiveapproach helpspeoplebetterpersonalizehealthservicesandmotivatesthemtoincreasephysicalactivity.InProceedings

of the 2015 ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp ‘15)(pp.

743-754).NewYork,NY:ACM.doi:10.1145/2750858.2807552

Li,I.,Dey,A.K.,&Forlizzi,J.(2010,April)AStage-BasedModelofPersonalInformaticsSystems.In

Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI ’10)(pp.557-566).

NewYork,NY:ACM.doi:10.1145/1753326.1753409

Li,I.,Dey,A.K.,&Forlizzi,J.(2012).Usingcontexttorevealfactorsthataffectphysicalactivity.ACMTrans. Comput.-Hum.Interact.,19(1).

Lops,P.,deGemmis,M.,&Semeraro,G.(2010).Content-basedRecommenderSystems:StateoftheArt andTrends.InF.Ricci,L.Rokach,B.Shapira,&P.B.Kantor(Eds.),Recommender Systems Handbook(pp. 73–105).NewYork,NY:Springer.

MacQueen,K.M.,McLellan-Lemal,E.,Bartholow,K.,&Milstein,B.(2008).Team-basedCodebook Development:Structure,Process,andAgreement.InG.Guest&K.M.MacQueen(Eds.),Handbook for

Team-Based Qualitative Research(pp.119–136).Lanham,MD:AltaMiraPress.

Marcengo,A.,&Rapp,A.(2013).VisualizationofHumanBehaviorData:TheQuantifiedSelf.InL.H.Huang &W.Huang(Eds.),Innovative approaches of data visualization and visual analytics(pp.236–265).Hershey, PA:IGIGlobal.

Marshall,B.P.,Cardon,P.A.,Poddar,A.,&Fontenot,R.(2013).DoesSampleSizeMatterinQualitative Research?AReviewofQualitativeInterviewsinIsResearch.Journal of Computer Information Systems,54(1), 11–22.doi:10.1080/08874417.2013.11645667

Marshall,M.N.(1996).Samplingforqualitativeresearch.Family Practice,13(6),522–525.doi:10.1093/ fampra/13.6.522PMID:9023528

Medynskiy,Y.,&Mynatt,E.(2010).Salud!:Anopeninfrastructurefordevelopinganddeployinghealth self-managementapplications.InProceedings of the 4th International Converence on Pervasive Computing

Technologies for Healthcare.doi:10.4108/ICST.PERVASIVEHEALTH2010.8897

Nielek,R.,&Wierzbicki,A.(2010,May).Emotionawaremobileapplication.InJ.S.Pan,S.M.Chen,&N. T.Nguyen(Eds.),Computational Collective Intelligence. Technologies and Applications. ICCCI 2010, LNCS (Vol.6422,pp.122–131).Berlin:Springer.doi:10.1007/978-3-642-16732-4_14

Ohlin,F.,&Olsson,C.M.(2015,March).IntelligentComputinginPersonalInformatics:KeyDesign Considerations.InProceedings of the 20th International Conference on Intelligent User Interfaces (IUI ‘15) (pp.263-274).NewYork,NY:ACM.doi:10.1145/2678025.2701378

Oppermann,R.(1994).Adaptive User Support: Ergonomic Design of Manually and Automatically Adaptable

Software.Hilldale,NJ:LawrenceErlbaum.

Perera,C.,Zaslavsky,A.,Christen,P.,&Georgakopoulos,D.(2014).Contextawarecomputingforthe internetofthings:Asurvey.IEEE Communications Surveys and Tutorials,16(1),414–454.doi:10.1109/ SURV.2013.042313.00197

Purpura,S.,Schwanda,V.,Williams,K.,Stubler,W.,&Sengers,P.(2011,May).Fit4life:thedesignof a persuasive technology promoting healthy behavior and ideal weight. In Proceedings of the SIGCHI

Conference on Human Factors in Computing Systems (CHI ‘11)(pp.423-432).NewYork,NY:ACM.

doi:10.1145/1978942.1979003

Rapp, A., & Cena, F. (2014). Self-monitoring and Technology: Challenges and Open Issues in PersonalInformatics.InProceedings of HCI International. Universal Access in Human-Computer

Interaction. Design for All and Accessibility

Practice(Vol.8516,pp.613-622).doi:10.1007/978-3-319-07509-9_58

Rapp,A.,&Cena,F.(2016).PersonalInformaticsforEverydayLife:HowUserswithoutPriorSelf-Tracking Experience Engage with Personal Data. International Journal of Human-Computer Studies, 94, 1–17. doi:10.1016/j.ijhcs.2016.05.006

Rapp,A.,Cena,F.,Kay,J.,Kummerfeld,B.,Hopfgartner,F.,Plumbaum,T.,&Gouveia,R.et al.(2016, September).Newfrontiersofquantifiedself2:goingbeyondnumbers.InProceedings of the 2016 ACM

International Joint Conference on Pervasive and Ubiquitous Computing: Adjunct (UbiComp ‘16)(pp.506-509).

NewYork,NY:ACM.doi:10.1145/2968219.2968331

Rapp, A., Cena, F., Kay, J., Kummerfeld, B., Hopfgartner, F., Plumbaum, T., & Larsen, J. E. (2015, September). New frontiers of Quantified Self: finding new ways for engaging users in collecting and using personal data. In Adjunct Proceedings of the 2015 ACM International Joint

Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2015 ACM International Symposium on Wearable Computers (UbiComp/ISWC’15 Adjunct)(pp.969-972).NewYork:ACM.

doi:10.1145/2800835.2807947 Rapp,A.,Marcengo,A.,Buriano,L.,Ruffo,G.,Lai,M.,&Cena,F.(inpress).Throughthelooking-glass: Quantifyingtheselfthroughauser-drivenpersonalinformaticssystem. Rapp,A.,&Tirassa,M.(2017).KnowThyself:AtheoryoftheselfforPersonalInformatics.Human-Computer Interaction,1–46.doi:10.1080/07370024.2017.1285704 Rooksby,J.,Rost,M.,Morrison,A.,&Chalmers,M.(2014,April).Personaltrackingaslivedinformatics.In

Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI ‘14)(pp.1163-1172).

NewYork,NY:ACM.

Schmidt, B., Benchea, S., Eichin, R., & Meurisch, C. (2015, September). Fitness tracker or digital personalcoach:howtopersonalizetraining.InAdjunct Proceedings of the 2015 ACM International Joint

Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2015 ACM International Symposium on Wearable Computers (UbiComp/ISWC’15 Adjunct)(pp.1063-1067).NewYork,NY:ACM.

doi:10.1145/2800835.2800961

Strauss,A.L.,&Corbin,J.M.(1990).Groundedtheoryresearch:Procedures,canons,andevaluativecriteria.

Qualitative Sociology,13(1),3–21.doi:10.1007/BF00988593

Tang,L.M.,&Kay,J.(2013,June).Lifelongusermodelingandmeta-cognitivescaffolding:Supportself monitoringoflongtermgoals.InWorkshop UMAP Proceedings.

Thaler,R.H.,&Sunstein,C.R.(2008).Nudge. Improving Decisions About. Health, Wealth, and Happiness. NewHaven,CT:YaleUniversityPress.

Federica Cena is Assistant Professor at the Department of Computer Science of the University of Torino. She is currently the head of Smart City Lab at the Center for Innovation for the Territory. She is working on user modeling and personalization, with a special focus on the implications of IoT for user modeling.

Amon Rapp holds a degree in Communication Sciences from the University of Torino, Italy. Since 2006 he’s been working in the Research and Trends department in Telecom Italia as a Human-Computer Interaction Researcher. At present, he’s a short-term researcher at Computer Science Department at University of Torino. His main research areas are about the use of game design elements in interactive systems, user research methodologies and Personal Informatics tools.

Silvia Likavec is a post-doctoral researcher at Computer Science Department, University of Torino, Italy. She obtained her PhD in Computer Science from University of Torino, Italy and École Normale Supérieure de Lyon, France in 2005 and her Master in Applied Mathematics from University of Novi Sad, Serbia in 2005. Her research activity is focused on the following: (i) modelling the user in social systems, in particular the design of the ontology-based user model and the propagation of user interests using this model; (ii) semantic similarity among ontological concepts; (iii) theoretical computer science, in particular type assignment systems, resource control calculi and semantics of programming languages. She is the author of more than 40 articles in prestigious international journals, monographs and peer-reviewed conferences and workshops and two text books for university level courses. Alessandro Marcengo is a Psychologist and Researcher with extensive expertise both in clinical domain and in Human Computer Interaction Research. As a researcher, he is author of numerous international publications in the field of HCI and participated in different European projects with various roles. His current research interest focuses on Quantified Self arising issues, on the impact of new technologies in terms of awareness and personal well-being and in the construction of meaning through Data Visualization systems. He is also chair and organizer of different tracks and workshops within different international conferences related with all the above Topics.

Tkalčič,M.,Burnik,U.,Odić,A.,Košir,A.,&Tasič,J.(2012).Emotion-aware recommender systems–a

framework and a case study. ICT Innovations(pp.141–150).Berlin:Springer.

Warshaw,J.,Matthews,T.,Whittaker,S.,Kau,C.,Bengualid,M.,&Smith,B.A.(2015,April).CananAlgorithm Knowthe“RealYou”?:UnderstandingPeople’sReactionstoHyper-personalAnalyticsSystems.InProceedings

of the 33rd Annual ACM Conference on Human Factors in Computing Systems (CHI ‘15)(pp.797-806).New

York,NY:ACM.doi:10.1145/2702123.2702274

Weiser,M.(1998).Thefutureofubiquitouscomputingoncampus.Communications of the ACM,41(1),41–42. doi:10.1145/268092.268108 ENdNoTES 1_{ https://www.google.com/fit/} 2_{ http://www.apple.com/it/ios/health/} 3_{ https://www.tictrac.com/} 4_{ https://www.headsuphealth.com} 5_{ https://www.beeminder.com} 6_{ https://www.myfitnesspal.com} 7_{ https://www.supertracker.usda.gov} 8_{ http://loseit.com}